The benchmark, and what it actually means for your account

The efficiency gap isn’t evenly distributed across flow types either, which matters more than the headline 41% number once you’re deciding where to focus. Back-in-stock flows average $9.14 in revenue per recipient. Abandoned cart flows average $3.65. A typical one-off campaign send averages $0.11. That’s not a small gap between your best-performing automation and your average email. It’s closer to two orders of magnitude, and it means the specific flow you haven’t built yet, or the one you built once and never touched again, is very likely costing you more in missed revenue than any campaign optimization you could run this month.

Your target share of revenue from flows depends on your stage, and Klaviyo’s own benchmark data breaks this out specifically: 25% to 35% of email revenue from flows is normal under $5 million in revenue, 40% to 50% in the $5 million to $20 million range, and 50% to 60% above $20 million, with the top decile of mature programs reaching 58% to 65%. If your flow revenue share is sitting meaningfully below your stage’s benchmark, that’s a specific, measurable gap, not a vague sense that email could be doing more.

One caveat worth taking seriously before you act on any of this: these are last-touch attribution numbers, not incrementality. Klaviyo’s default five-day attribution window tends to overstate how much of a purchase email genuinely caused versus simply touched last before a sale that would have happened anyway. Use the relative gaps between flow types, back-in-stock dramatically outperforming a standard campaign, to find your weak spots. Don’t treat the specific dollar figures as precise accounting truth.

The tool built specifically to run this audit for you

Klaviyo’s Flows Analyst feature, built directly into the platform, benchmarks your own welcome, browse abandonment, cart abandonment, and win-back flows against what’s actually working across Klaviyo’s full brand dataset, then surfaces specific recommendations rather than leaving you to compare your own numbers manually against a benchmark blog post. Instead of guessing whether your 22% cart abandonment open rate is good or mediocre, the tool tells you directly where you stand relative to brands running the same flow type at your approximate scale.

This solves a real, specific problem most teams have with flow optimization: knowing a benchmark exists is different from knowing exactly where your own account falls short of it. A blog post telling you the industry average abandoned cart click rate is 9.5% is useful context. A tool that pulls your actual flow’s click rate and tells you directly whether you’re above or below that line, and which specific flow needs attention first, is the difference between having information and being able to act on it immediately.

How to actually run this audit this week

Start with your foundational five: welcome, abandoned checkout, post-purchase, win-back, and browse abandonment. Multiple independent benchmark sources converge on these as the flows every brand should have live before building anything more advanced, and they’re also the flows Klaviyo’s Flows Analyst is built to evaluate directly. If any of these five isn’t currently live in your account, that’s a bigger gap than any optimization tweak to a flow you already have running.

For each flow that is live, check two numbers specifically: entry-email conversion rate, meaning orders attributed to just the first email divided by how many people received it, and full-flow conversion rate, the total orders from the entire flow divided by total recipients across every message in it. These measure genuinely different things, and confusing them is a common audit mistake. A four-email flow will always show a lower full-flow conversion rate than a one-email flow at the same order volume, simply because the denominator is larger. Compare your flow against benchmarks using the same metric definition, not different ones.

Prioritize your fixes by where the dollar gap is largest, not by which flow feels easiest to edit. Given that back-in-stock flows average $9.14 per recipient against a campaign’s $0.11, a back-in-stock flow that doesn’t exist yet, or exists but is thin, is a bigger opportunity than polishing subject lines on a campaign that was never going to move the needle much regardless of how well it’s written.

If your welcome flow’s discount offer is a token 5% “welcome gesture,” treat that specifically as a fixable pattern rather than a minor detail. Real, effective welcome flows tend to run 15% to 20% off the first order, timed tightly, first email within minutes of signup rather than immediately, second email within one to two days rather than a week later. A shallow discount that doesn’t meaningfully cover a new subscriber’s actual purchase hesitation is a common, specific reason a welcome flow underperforms its potential despite being technically live.

Your action list

• Open Flows Analyst inside Klaviyo and run it against your five foundational flows, welcome, abandoned checkout, post-purchase, win-back, and browse abandonment. Note which ones fall below the tool’s benchmark comparison, not just which ones feel like they could be better.

• Calculate your current flow revenue share as a percentage of total email revenue, and compare it against your stage’s benchmark, 25 to 35% under $5M, 40 to 50% at $5M to $20M, 50 to 60% above $20M. If you’re meaningfully below your stage’s range, that’s your priority, not a new campaign idea.

• Check whether you have a back-in-stock flow live at all. Given it’s the single highest revenue-per-recipient flow type in Klaviyo’s own benchmark data, at $9.14 versus a campaign’s $0.11, a missing or thin back-in-stock flow is very likely the biggest single gap in your entire email program right now.

Sources: Klaviyo, "How to Use Klaviyo Benchmarks Feature: 3 Industry Use Cases"; Klaviyo, "What's New: Product Updates and Releases"; Eightx, "Klaviyo flow revenue benchmarks by brand stage" (June 2026); Solo Media Group, "Klaviyo for Shopify: The 5 Flows That Drive 80% of Revenue"; BS&Co, "Klaviyo Flow Benchmarks: Real Conversion Rates" (May 2026); 624 Agency, "Klaviyo flows every ecommerce brand needs in 2026."

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What actually changed, and why it’s easy to miss

For any budget-limited campaign running Target CPA, Target ROAS, or Target CPC for Demand Gen, Smart Bidding has historically been allowed to hunt for the cheapest available conversions within your budget, even when that meant beating your stated target by a wide margin. A $10 Target CPA campaign quietly delivering conversions at $5 wasn’t an error. It was the algorithm finding efficient auctions and taking them, and if you never noticed the gap between your target and your actual performance, there was no reason to touch anything. The campaign was working, arguably better than you’d asked it to.

As of August 17, Google changed that behavior directly. Campaigns marked “Limited by budget” using a target-based bid strategy now optimize consistently toward the exact number you typed in, not toward whatever the algorithm can find below it. Google’s own framing treats this as a predictability fix, and on its own terms, it’s a reasonable one: an advertiser who increases budget on an over-performing campaign has historically seen inconsistent results, because the algorithm’s behavior at $5 actual CPA doesn’t scale predictably once you add spend. Google wants budget increases to produce predictable outcomes. The tradeoff is that any campaign happily beating its target gets pulled back to that target whether you wanted that or not.

Multiple sources covering this change flag the same specific detail worth understanding precisely: this doesn’t touch every campaign type. It applies to Search, Shopping, Performance Max, Demand Gen, and Travel campaigns using Target CPA, Target ROAS, or Target CPC. It does not apply to Maximize Conversions or Maximize Conversion Value campaigns without a set target, manual CPC, target impression share, or target CPM. If your account runs a mix of strategies, only a specific subset of your campaigns is actually exposed to this change, which means a blanket account-wide reaction is the wrong response. The right response is identifying exactly which campaigns qualify.

Why this is worse than it sounds for anyone who set a target once and moved on

The instinct might be to assume this only affects advertisers who deliberately gamed their targets low to force aggressive bidding. That’s not the real exposure. The far more common situation, and the one multiple sources covering this change specifically call out, is a target that was set reasonably at some point in the past and simply never revisited once the campaign started outperforming it. Nobody goes back and lowers a Target CPA from $10 to $5 just because the campaign happens to be delivering at $5. Why would you. The campaign is working. That’s precisely the account Google’s change now quietly penalizes, not through any error or account issue, but through a deliberate behavioral change that treats your old, un-updated target as the literal instruction it always technically was.

For campaigns running inside a shared budget or portfolio bid strategy, the exposure compounds. If a shared budget pool includes even one campaign that’s technically budget-limited, the impact spreads across every campaign drawing from that pool, whether or not each individual campaign looks budget-constrained on its own. That’s a meaningfully wider blast radius than “check your budget-limited campaigns individually” suggests, and it’s the specific reason multiple agency sources covering this change flag portfolio strategies as needing a different, more careful audit than single-campaign accounts.

The tool Google built specifically for this, and how to actually use it

Google shipped a Bid Target Adjustment Tool inside Google Ads starting July 6, six weeks ahead of the August 17 change, specifically to surface which of your campaigns are exposed and give you a direct way to respond before the shift happens. If you haven’t opened it yet, the tool identifies campaigns where recent actual performance has been meaningfully better than the stated target, and offers a direct option to update your target to match that recent performance, preserving roughly the outcome you’d been getting rather than losing it to this change by default.

The tool’s core function solves the specific problem this change creates: you don’t have to manually dig through months of performance history for every campaign to figure out which targets are stale. It surfaces the gap for you. What it doesn’t do is decide for you whether that gap was a genuine favorable outcome worth locking in, or whether your original target actually reflected what you wanted and the campaign’s recent over-performance was itself the anomaly worth investigating before you enshrine it as your new normal.

What to check this week

Open the Bid Target Adjustment Tool inside Google Ads and review every campaign it surfaces, not just your highest-spend ones. A smaller campaign that’s been quietly overdelivering is just as exposed as a large one, and it’s easy to deprioritize checking it simply because the dollar amounts involved feel less urgent.

For each flagged campaign, make an actual decision rather than defaulting to “apply” without thinking about it. If a campaign’s recent $5 actual CPA against a $10 target genuinely reflects sustainable, repeatable performance, updating your target to match protects the outcome you’ve been getting. If that $5 CPA was itself unusual, a seasonal dip, a temporary competitive gap, a recent creative that happened to convert unusually well, locking in that number as your new target may set an unrealistic bar the campaign can’t consistently hit going forward.

If you run shared budgets or portfolio bid strategies, treat this as a portfolio-level audit, not a campaign-by-campaign one. Identify whether any single campaign inside a shared pool is budget-limited, since that status can pull the entire pool’s behavior along with it even if most of the individual campaigns inside it don’t look constrained in isolation.

Recheck this in 30 days regardless of what you find today. Because Google won’t automatically adjust your targets or your budgets going forward, drift between target and actual performance can reopen the same gap this change just closed, and there’s no future automatic correction coming next time. This is now a setting that needs a recurring check, not a one-time fix.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

How ButcherBox killed its “monolith” email strategy and built a retention team around individualized data instead

ButcherBox, the meat and seafood subscription service that reported more than $600 million in revenue in 2022, built an entire internal department dedicated specifically to churn and retention, a structural commitment most subscription brands never make. Chief Commercial Officer Reba Hatcher told Modern Retail at SubSummit that the shift away from what her team internally called “monolith behavior,” sending every customer the identical email with the identical offer at the identical time, was central to sustaining growth even as consumer sentiment softened.

The numbers: ButcherBox declined to share exact current sales figures, but confirmed the retention strategy was directly credited with continuing to grow the business through a period of lower consumer spending confidence. The company launched its loyalty program, Sizzle Society, in late 2024, offering tiered credits and perks, including a virtual Q&A series with the CEO for members at the top “executive chef” tier.

Why it worked: Hatcher was specific about the actual mechanism, not just the philosophy behind it. The company moved from blanket, identical messaging toward individualized data, factoring in what specific products a customer actually buys and where they live, to shape what that customer sees. That’s a meaningfully different operational commitment than simply having a loyalty program, since a tiered rewards system layered on top of identical mass messaging still leaves the underlying communication generic. ButcherBox’s real shift was upstream of the loyalty program itself: building the data and personalization infrastructure needed to make every subsequent touchpoint, loyalty perks included, actually specific to the individual customer receiving it.

How to replicate it:

Audit your own retention emails for genuine “monolith behavior” before assuming a loyalty program alone will fix churn. If every subscriber, regardless of purchase history or location, currently receives the identical win-back offer at the identical interval, that’s the specific pattern ButcherBox identified and dismantled, not a loyalty tier structure sitting on top of it.

Use purchase history and basic geographic data as your starting personalization layer before investing in anything more sophisticated. ButcherBox’s example, factoring in what customers actually buy and where they live, is a relatively accessible starting point most subscription and DTC brands already have the raw data for, even without dedicated retention headcount.

If retention matters enough to your business model that ButcherBox built an entire department around it, treat that as a signal about resourcing, not just tactics. A loyalty program bolted onto a small team’s existing workload competes for attention against every other priority. A brand serious about retention as a growth lever, the way ButcherBox describes it, may need to treat it as a distinct function with its own ownership, not a side project inside marketing.

Sources: Modern Retail, "How subscription brands Babbel, BattlBox and ButcherBox are fighting for retention as people tighten their wallets," reporting by Mitchell Parton, featuring Reba Hatcher, Chief Commercial Officer at ButcherBox, Google Ads Help, "Frequently asked questions about changes to Target-based bid strategies" (support.google.com); Google Ads Help, "Changes to target based bid strategies" (support.google.com); Search Engine Roundtable, "Google Ads Changes With Bidding For Campaigns Limited By Budget" (July 2026); Search Engine Journal, "Google Is Ending Target Overperformance: What to Fix Before August 17" (August 2026); Optmyzr, "Google's August 17 Bidding Change: What Advertisers Need to Do Now"; Keyweo, "Google Ads update, August 17, 2026" (July 27, 2026); 4M Digital Consulting, "Google Ads' August 17 Bidding Change: What PPC Managers Must Do"; Modern Retail, "How subscription brands Babbel, BattlBox and ButcherBox are fighting for retention as people tighten their wallets," Mitchell Parton.

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually changing today

Shopify has been retiring checkout.liquid, the old system that let merchants paste raw HTML, CSS, and JavaScript directly into checkout, in phases for two years. The core checkout pages, Information, Shipping, Payment, already moved off the legacy system back in August 2024. Plus stores hit their own Thank You and Order Status deadline on August 28, 2025. Today is the second wave: every non-Plus plan, Basic, Shopify, and Advanced, regardless of revenue or store size, hits the same cutoff Plus stores already went through a year ago.

The mechanism matters more than the date itself. This isn’t a warning followed by a grace period. Shopify auto-upgrades stores that haven’t migrated and removes the legacy customizations as part of that automatic process. There’s no opt-out and, per multiple independent guides on the migration, the legacy customizations are not preserved anywhere you can retrieve them afterward. Whatever was sitting in your Additional Scripts field becomes frozen and then removed, not archived somewhere you can copy from later.

The replacement framework is Checkout Extensibility: Checkout UI Extensions, Shopify Functions, and Shopify Pixels, sandboxed, structured tools that replace the old open-ended script injection. For non-Plus merchants specifically, the customizable surface is narrower than what Plus stores get. Checkout UI Extensions for the core checkout steps, Information, Shipping, Payment, are Plus-only. Non-Plus merchants can customize the Thank You and Order Status pages, which is exactly where most tracking pixels and post-purchase apps have historically lived anyway.

Why the silent failure is the actual danger here

A hard error is easy to catch. A page that won’t load gets noticed within hours because customers complain or a support ticket comes in. This is the opposite failure mode, and it’s the reason this deadline deserves more attention than a typical platform update. Orders keep processing. Checkout keeps working from a customer’s perspective. The only thing that breaks is the invisible layer of tracking and attribution sitting on top of a transaction that otherwise completes normally. You don’t get an error message. You get a slow, quiet divergence between your actual sales and what your ad platforms report, and that gap is the kind of thing that only becomes obvious weeks later when someone finally asks why ROAS looks worse this month for no clear reason.

There’s a second layer worth understanding, since it explains why some stores are already partially affected without realizing it. Shopify stopped passing personally identifiable information to Additional Scripts on legacy pages back in August 2025, a full year before today’s hard cutoff. That means stores still running the old system have likely already been losing attribution quality gradually for the past year, not just facing a clean break starting today. If your conversion tracking has felt quietly worse over the past several months and you assumed it was iOS privacy changes or platform-side measurement drift, part of that decline may simply have been this earlier, less-publicized change already in effect.

What to actually check today, in order

Open Settings, then Checkout, and look at the Additional Scripts field directly. Don’t rely on memory of what you or an agency put there months or years ago. Document everything currently sitting in that field: which pixel, which platform, what it’s tracking, and which app or team originally installed it. This single audit, per multiple sources covering the migration, takes about an hour and is the entire foundation for everything that follows.

Check whether you’re actually still on the legacy system or already auto-upgraded. Auto-upgrades have reportedly been rolling out gradually since earlier this year for some stores, meaning your store may already be running on the new Checkout Extensibility system without you having deliberately migrated anything. If that’s the case, your actual task shifts from migration to verification: confirming your replacement tracking is live and firing correctly, not building it from scratch.

For each script or pixel you find in the audit, identify its official replacement path. Major platforms, Google, Meta, TikTok, have native channel apps built specifically for the Web Pixels API that replace manual pixel installation. Custom or less common scripts, affiliate tracking, specific post-purchase apps, need individual verification with whichever vendor built them, since not every legacy integration has an automatic equivalent waiting on the other side.

Place a real test order today, not after you assume the migration is complete. Check your ad platform’s events manager or conversion dashboard to confirm the test purchase actually registered. This is the step most guides on this migration flag as the one people skip, assuming an app update or a settings toggle handled everything, only to discover weeks later that a specific pixel never reconnected.

Your action list

• Open Settings, then Checkout today and document everything currently in your Additional Scripts field, along with your Thank You and Order Status page customizations, before assuming anything has already been handled for you.

• Install the native channel apps for whichever major ad platforms you run, Google, Meta, TikTok, if you haven’t already, since these replace manual pixel scripts through the supported Web Pixels API rather than the deprecated injection method.

• Place one real test order this week and manually verify inside each ad platform’s own dashboard that the purchase event actually registered. Don’t consider this migration complete until you’ve confirmed it with a live transaction, not just a settings screen that looks correctly configured.

Sources: Shopify Help Center, checkout extensibility and Web Pixels API documentation; Digital Applied, “Shopify Checkout Deadline: 12 Days to Migrate” (August 2026); Codilar, “Shopify Checkout Extensibility August 26 Deadline: Important for Non-Plus Merchants” (July 2026); Flatline Agency, “Shopify checkout extensibility: what non-Plus stores need to do before August 2026”; Biscuits Bundles, “Shopify Checkout Extensibility for Non-Plus Stores: What Breaks on August 26, 2026”; Kaspian Fuad, “Shopify Additional Scripts Deprecated (Aug 26 2026)”; Revize, “Shopify Checkout Extensibility 2026: June 30 Deadline Guide” and “Shopify’s August 26 Checkout Deadline: What Breaks for Non-Plus Stores”; AuditIQ, “Shopify’s August 26 checkout deprecation: How to prepare.”

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

The problem, specifically

Ad platforms bill you based on delivery, impressions served, clicks recorded, conversions attributed, and when delivery breaks, the billing usually doesn’t stop cleanly alongside it. During the July 19 incident specifically, multiple advertisers reported being unable to pause active campaigns through the interface while the outage was ongoing, meaning spend continued against a platform that wasn’t reliably delivering or reporting during that window. A campaign that looked like it underperformed on July 19 might not have underperformed at all. It might have been spending into a broken pipe for two and a half hours.

There’s a second, quieter problem sitting next to the confirmed outages. Reports of publishing errors on July 21 and frozen-spend issues on July 22 circulated among advertisers, but neither has a matching entry on Meta’s own status page, which means they sit in a gray zone, real enough to generate complaints, unconfirmed enough that Meta hasn’t officially acknowledged them. That distinction matters directly for how you approach a credit request: a confirmed, dated, timed incident on Meta’s status page is a strong claim. An unconfirmed pattern of complaints from the same week is a weaker one, and treating the two the same way in a support request will get the weaker one dismissed.

What to actually check, and how

Pull your Meta Ads Manager reporting for both confirmed windows: July 16, roughly two hours, and July 19, two hours and thirty-two minutes. Look specifically for anomalies: spend that continued at a normal or elevated rate while conversions, clicks, or reported delivery dropped sharply relative to the hours immediately before and after. That gap, spend continuing while performance craters for a specific, narrow window, is the signal worth investigating rather than dismissing as a bad day.

If you find that gap, screenshot the delivery data for that specific window before doing anything else. Delivery and reporting data in ad platforms isn’t permanent in the way you might assume, and having a timestamped record of what your dashboard actually showed during a confirmed outage window is the difference between a credit request with evidence attached and one that’s just an assertion.

Annotate both confirmed dates directly inside whatever reporting view or dashboard you use regularly, not just in a separate notes document. If a campaign’s weekly or monthly performance looks soft and July 16 or July 19 falls inside that window, you want that annotation visible at the moment you’re evaluating the campaign, not buried somewhere you’d only find it if you already suspected an outage was the cause. This is exactly the kind of context that gets lost when a monthly report gets pulled by someone who wasn’t watching the account in real time during the incident itself.

File the actual credit request through Meta’s ads support, referencing the specific confirmed incident, its exact duration, and the screenshotted anomaly in your own account’s delivery data during that window. A vague request citing “issues in July” is far weaker than one that says, specifically, spend continued at your normal daily rate between a stated start and end time on a confirmed outage date while conversions dropped by a specific, calculable percentage relative to your typical performance in that same daypart.

Why this is worth the twenty minutes it takes

The instinct to skip this is understandable. Two hours here, two and a half hours there, doesn’t feel like enough to chase down against the time it takes to pull reports and file a request. But run the actual math against your own daily spend before deciding it isn’t worth it. A brand spending $2,000 a day on Meta has roughly $170 to $220 sitting inside a 2-to-2.5-hour window, and that’s before accounting for the fact that spend during a broken-delivery outage is disproportionately likely to have been wasted entirely rather than simply delivering at a normal, if unmeasured, rate. For a brand spending $10,000 or more per day, the same math scales to real money sitting in an unfiled credit request most teams never submit.

There’s a broader habit worth building here beyond this specific July incident. Platform outages happen with some regularity, and the operators who consistently recover credit for them are the ones who’ve built a standing habit of checking status pages against their own spend data after any period of unexplained performance softness, not the ones who only think to check after reading about an outage somewhere well after the fact.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

How Once Upon a Farm learned that shelf placement, not just shelf presence, decides whether a merchandising bet actually works

Once Upon a Farm, the organic baby food and kids’ snack brand founded in 2015, built its retail growth around an unusual bet: placing branded refrigerated coolers directly in grocery stores’ baby food aisles, rather than accepting standard shelf space next to shelf-stable competitors. CMO Katie Marston told Modern Retail Podcast the company learned early that where exactly a retailer agreed to put the cooler mattered as much as whether they agreed to carry it at all.

The numbers: the company ended 2025 with more than 3,400 coolers placed in grocery stores across the US and is targeting more than 5,000 by the end of 2026. Once Upon a Farm went public in February 2026, and its Q2 2026 earnings, released the same week as Marston’s interview, showed net sales of $85.4 million, up 42.3% year over year. Marston was direct that the cooler strategy is a major part of that growth story, not a peripheral marketing tactic.

Why it worked: the mechanism only becomes visible when you compare two different retail partnerships directly. When the brand first tested the cooler concept with Target, the fridges were placed at aisle end-caps, visible in passing but easy for a parent to walk past without noticing. When Kroger came on board, the fridges went directly inside the baby aisle itself, positioned exactly where a parent shopping that category was already standing. Same product, same cooler concept, same underlying pitch to the retailer, meaningfully different result depending on where inside the store the fixture actually landed. A retailer saying yes to a new merchandising format isn't the same as that format actually working, and the gap between those two things is entirely about the specific physical placement negotiated after the yes.

How to replicate it:

If you’re negotiating any kind of in-store placement, whether a display, a cooler, an endcap, or a fixture, treat the exact physical location as a separate negotiation from getting the retailer to agree to the concept at all. Getting to yes and getting a placement that actually performs are two different wins, and only one of them shows up in a signed agreement.

Ask directly where else in the store a similar fixture or product from a different brand already performs well, and push for placement that matches that pattern rather than accepting whatever spot a retailer offers by default. Once Upon a Farm’s Kroger placement worked specifically because it matched where parents were already standing, not just where there was open floor space.

If a retail test underperforms, diagnose placement specifically before concluding the format itself doesn’t work. Once Upon a Farm’s Target test could easily have been read as evidence that coolers don’t move parents to buy. The real lesson was narrower and more fixable: that specific placement didn’t work, and a different one, tested with a different retailer, did.

Track your retail partnerships individually rather than as a single aggregated wholesale channel number. The difference between Target’s end-cap result and Kroger’s in-aisle result would be invisible in a blended wholesale revenue figure, and it’s exactly the kind of comparison that reveals what’s actually driving performance versus what’s just adding SKU count to a channel.

Your action list

• Pull your Meta Ads Manager delivery data for July 16 and July 19 specifically, and check whether spend continued at a normal rate while conversions or delivery dropped sharply during Meta’s confirmed outage windows on those dates.

• If you find an anomaly, screenshot it immediately along with the specific timestamps, and file a credit request through Meta’s ads support referencing the confirmed incident directly rather than describing the issue generally.

• Set up a standing habit going forward: whenever a campaign’s performance looks unexplainably soft for a specific day or window, check Meta’s status page for that exact date before assuming the problem was creative, targeting, or budget. The platform’s own confirmed incident history is public and checkable, and it’s the fastest way to rule out, or confirm, whether a bad day was actually your campaign’s fault at all.

Sources: Meta Status page (metastatus.com), confirmed Ads Delivery incident logs for July 16 and July 19, 2026; AdMake AI Blog, "Meta Ads Updates: August 2026 Changelog for Media Buyers"; AdsUploader, "Meta Ads Updates (August 2026): What's Changing and What to Do.", Modern Retail Podcast, "How Once Upon a Farm is shaking up the baby aisle with coolers," featuring CMO Katie Marston (August 2026); FoodNavigator-USA, "Kroger and Once Upon a Farm partner to test coolers in baby aisle"

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

The tool: Klaviyo Extended ID

Extended ID is a first-party identity feature that extends how long Klaviyo can recognize a returning visitor, from the standard 7-day window Safari’s Intelligent Tracking Prevention forces on most browsers, up to a full year. Mechanically, it works by reading common deterministic identifiers already stored in a visitor’s browser, cookies like _ga, _ttp, and _uetvid that Google Analytics, TikTok, and Microsoft’s ad tools already set, and using them to reconnect a returning visitor to their existing Klaviyo profile even after their original tracking cookie has expired.

The practical effect: a customer who visited three weeks ago, got cookied by Safari’s 7-day limit, and effectively vanished from Klaviyo’s view, can now be recognized again the moment they return, without you doing anything differently in how you run ads or emails. Klaviyo’s own documentation cites a 5 to 10% lift in visitor identification with zero additional setup required once it’s live, which directly means more people entering your abandoned cart, browse abandonment, and win-back flows who would otherwise have been triggering nothing at all.

Why this specific setting matters more than another flow

Most optimization advice defaults to building a new flow or writing better subject lines. Extended ID sits underneath all of that. Every flow you’ve already built, welcome series, abandoned cart, browse abandonment, win-back, only fires for people Klaviyo can actually identify. If your identification rate is quietly capped by Safari’s 7-day cookie limit, you don’t have a flow-copy problem. You have a visibility problem, and no amount of better subject lines fixes a flow that never triggers because the platform lost track of the person it was supposed to trigger for.

This matters more today than it would have two years ago, specifically because of how much traffic now arrives from a first click that doesn’t convert immediately. A shopper who discovers you through a TikTok video, an Instagram comment reply, or an organic search result rarely buys on that first visit. They come back later, often on a different device or after their original cookie window has already closed. Extended ID is built directly for that exact gap between first discovery and eventual purchase, which is a longer gap for most brands today than it was when 7-day cookie windows were originally considered a reasonable tracking horizon.

What it doesn’t do, so you’re not expecting the wrong thing

Extended ID re-identifies people who already have a Klaviyo profile. It does not create a new profile for a stranger who’s never interacted with your brand, and it doesn’t use device fingerprinting or any cross-site tracking method. If someone has never signed up, clicked an email link, or gone through checkout, Extended ID has nothing to reconnect them to. It’s a retention tool for identity you’ve already captured once, not a way to identify anonymous traffic from scratch. Setting accurate expectations here matters, because the lift you’ll see is concentrated in returning visitors specifically, not in your total traffic count.

How to switch it on

Extended ID lives in your Klaviyo account settings and is available on every paid Klaviyo plan already, meaning if you’re already paying for Klaviyo, there’s no additional cost to enable this. The setup itself is a toggle, not a development project, though Klaviyo’s own guidance is specific that you should update your cookie consent notice before turning it on, since Extended ID holds identifying information significantly longer than a standard tracking cookie, and your customers deserve to know that.

Once it’s live, here’s the specific way to check whether it’s actually working, rather than just assuming a toggle did something: go to Analytics, then Metrics, then Active on Site. Add a filter for extended_id. Events captured through Extended ID show a value of 1.0. Events captured through Klaviyo’s standard cookie tracking show 0.0. That comparison, run over a two to three week window after activation, is your real before-and-after: how many additional identified visits Extended ID is generating that your standard tracking would have missed entirely.

If you want to confirm your basic identification setup is functioning correctly before you even get to Extended ID specifically, there’s a simple test worth running first. Append “?klaviyo_identify=youremail@example.com“ to your own store’s URL, using an email you can search for. Reload the page, then search that email inside Klaviyo. If a profile shows the identification event, your baseline tracking is working correctly, which matters because Extended ID improves on a foundation that needs to already be functioning, not one that’s broken from the start.

Your action list

Turn on Extended ID today, and update your cookie notice at the same time rather than treating that as a follow-up task. Then set a calendar reminder for three weeks out, not sooner, since Extended ID's value compounds specifically over the kind of longer window a standard 7-day cookie would have already lost. At that three-week mark, pull the Active on Site report filtered by extended_id, and separately check whether your abandoned cart and browse abandonment flow trigger volume has moved. If your flows are seeing more entries without any change to your traffic sources or ad spend, that's Extended ID doing exactly the job it's built for: recovering identity your platform used to lose on its own.

Sources: Klaviyo Help Center, "How to set up extended ID cookie tracking"; Klaviyo Help Center, "About cookies in Klaviyo"; Klaviyo Help Center, "Understanding cookies in Klaviyo"; Klaviyo, "Capture More Data with Klaviyo Extended ID & Shopify Pixel Sync"; Klaviyo, "What Is Cookie Consent?"; Klaviyo, "All Klaviyo Product Updates and Releases."

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually happening

The deal itself closed cleanly after UK regulatory approval came through July 15, following a Competition and Markets Authority merger inquiry that had been running since June. eBay paid $1.2 billion in base purchase price plus roughly $200 million in net adjustments and interest, landing at the $1.4 billion final figure. Depop keeps its brand, its platform, its 56.3 million registered users, and its existing CEO, Peter Semple, operating as what eBay calls a complementary business rather than being folded into eBay’s core marketplace. eBay’s own framing on close was explicit about where it sees the value: combining Depop’s category leadership with eBay’s shipping, personalization, compliance, and trusted-services infrastructure, aimed at accelerating growth eBay believes Depop’s standalone scale couldn’t reach on its own.

Etsy’s side of the announcement is where the more interesting story sits. CEO Kruti Patel Goyal framed the sale as enabling Etsy to focus exclusively on growing its core marketplace, and the company confirmed proceeds would go toward general corporate purposes consistent with its April capital allocation strategy, including accelerating its share buyback program. That’s a company choosing to convert a strategic asset into direct cash return to shareholders rather than continuing to fund a second platform’s growth. The timing compounds the signal: Etsy’s 12% workforce reduction landed in the same reporting window as the Depop close, and CEO Goyal was explicit in public comments that neither cost-cutting nor AI drove the restructuring on its own, while still acknowledging AI is changing how the company works. Whatever the precise internal reasoning, a company selling a major acquisition below its purchase price and cutting more than one in ten jobs in the same month is a company narrowing its focus under real pressure, not one making a leisurely portfolio adjustment.

The numbers underneath the deal explain both sides of the decision. Depop generated roughly $1 billion in annual gross merchandise volume in 2025. eBay’s own Fashion category alone already generates more than $10 billion in GMV annually, meaning Depop represents a real but genuinely modest addition to eBay’s existing fashion business in pure volume terms, its actual value to eBay is concentrated in what it brings that eBay doesn’t already have: a mobile-first platform where close to 90% of active buyers are under 34, a demographic eBay’s core marketplace has historically struggled to reach.

Why this deal matters beyond two public companies’ M&A calendars

There’s a competitive detail sitting just behind this transaction that changes how it should be read, and most of the coverage treating this as a straightforward eBay-strengthens-its-fashion-category story is missing it. Vinted, the Lithuanian resale marketplace, reported 2025 GMV of €10.8 billion, up 47% year over year, and entered the US market in January 2026. That’s not a distant competitor. That’s a fast-growing, well-capitalized platform moving directly into the exact demographic and category Depop occupies, at the exact moment ownership of Depop is changing hands. eBay isn’t just acquiring a fashion resale platform in the abstract. It’s acquiring one in the middle of a live competitive fight against an entrant growing at nearly 50% a year, and the real test of this deal isn’t whether Depop’s GMV number looks respectable next to eBay’s existing Fashion category. It’s whether eBay’s infrastructure and capital can help Depop hold ground against Vinted faster than Depop could have done it alone under Etsy’s ownership.

That reframes what actually went wrong for Etsy, if anything did. Etsy’s own public reasoning, wanting to focus exclusively on its core marketplace, is a legitimate strategic position independent of Depop’s performance. But the fact that Etsy is exiting at a loss on the purchase price while a faster-growing competitor enters the same market Depop operates in is worth sitting with as a caution about a specific pattern: acquiring a platform in a hot category during a valuation peak, then being unable or unwilling to fund the ongoing investment needed to defend that category once a well-funded new entrant shows up. Etsy bought Depop in 2021 near the top of pandemic-era resale enthusiasm. It’s exiting in 2026 into a market that’s gotten meaningfully more competitive since, not less, and the exit price reflects that shift as much as it reflects anything specific to Depop’s own execution.

For any operator running a marketplace presence, a multi-platform strategy, or evaluating whether to expand into an adjacent channel through acquisition or partnership, the transferable lesson isn’t about resale fashion specifically. It’s about the gap between acquiring category leadership and having the sustained capital and strategic focus to actually defend it once you have it. A platform that looks like a smart category bet at acquisition can become a distraction from your core business within a few years if a faster-moving competitor enters the same space and you’re not positioned, or willing, to keep funding the fight.

What to actually take from this if you’re not running a billion-dollar marketplace

If you operate on Depop, or on any resale or C2C marketplace generally, watch the ownership transition closely over the next two quarters rather than assuming brand continuity promises mean nothing changes operationally. eBay has publicly committed to preserving Depop’s brand, platform, and community, which is the standard language in any acquisition of this kind. What matters more for a seller’s actual business is whether fee structures, algorithm changes, or cross-listing requirements shift as eBay integrates the platform into its own infrastructure, since “complementary business” status doesn’t guarantee the underlying mechanics stay static.

If you’re evaluating any acquisition or major platform partnership for your own business, treat Etsy’s experience as a real, recent case study in the gap between acquiring a growth category and sustaining the investment required to defend it once acquired. A category that looks attractive at the moment of acquisition can look meaningfully different five years later if a well-funded competitor enters in the interim, and the willingness to keep funding the fight, not just the initial purchase decision, is what actually determines whether the acquisition pays off.

If you compete in resale, secondhand, or any C2C category at all, Vinted’s 47% GMV growth and January 2026 US entry is worth tracking directly regardless of whether you have any relationship with Depop, eBay, or Etsy. A competitor growing that fast in a category adjacent to your own is a market-shape signal, not just a data point about one specific platform transaction.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

How Vinted is funding a slower, more expensive US entry than most platforms would risk, and telling investors exactly why

Vinted, the Lithuanian secondhand marketplace, launched its US expansion in January 2026 after maintaining only a dormant presence in the market since 2013. CEO Adam Jay told CNBC at London Tech Week in June that the shift toward secondhand consumption is a fundamental change that’s here to stay, but was direct about something most executives soften in public: building the American market could take years, not quarters, and the company is choosing to fund that timeline deliberately rather than rush it.

The numbers: Vinted’s 2025 revenue rose 38% to €1.1 billion, and GMV climbed 47% to €10.8 billion. Net profit fell 19% to €62 million over the same period, a decline the company attributes directly to heavy investment in its logistics arm, Vinted Go, its payments infrastructure, Vinted Pay, and geographic expansion into new markets including the US. The company’s April 2026 secondary share sale valued it at €8 billion, a 60% premium to its October 2024 valuation, led by EQT Growth with Schroders Capital and Teachers’ Venture Growth joining as new investors. Vinted’s own January 2026 research found 50% of Americans wear half their closet or less, and 42% are unaware their old clothing has resale value at all, the specific demand gap the company is targeting with its US entry.

Why it worked: Jay’s public framing treats declining net profit not as a warning sign to manage around quietly, but as the direct, disclosed cost of a specific, named strategic choice: building shipping and payments infrastructure before chasing volume, rather than chasing volume first and building infrastructure to catch up later. That sequencing matters directly against the eBay-Depop story above. Etsy owned Depop for five years without, on the available evidence, building the kind of dedicated logistics and payments infrastructure investment Vinted is now making as it enters the same broad market. Vinted’s willingness to show investors a falling profit number alongside an explicit explanation of what that decline is buying, rather than obscuring it, is itself a signal of confidence that the infrastructure investment will pay off over a longer horizon than quarterly reporting normally rewards.

How to replicate it:

If you’re funding a genuine expansion, whether a new channel, a new market, or a new product line, decide explicitly whether you’re investing in infrastructure first or chasing volume first, and communicate that choice clearly to whoever you’re accountable to, rather than letting a declining margin read as an unexplained problem.

Treat a competitor’s public timeline honesty as real information. Jay telling CNBC directly that US success could take years is a specific, checkable claim you can hold the company to later, and it’s also a signal about how much patience and capital Vinted itself believes this specific market requires, useful context if you’re assessing how seriously to treat them as a competitive threat right now versus over the next two to three years.

If you’re evaluating your own multi-year category investment, run the version of Vinted’s own research question against your own market: what percentage of your addressable customers don’t yet know the actual value of what you’re offering. Vinted’s stat, 42% of Americans unaware of their own clothing’s resale value, is a demand-education gap the company is explicitly building its US entry around, and most categories have some equivalent gap worth quantifying before assuming demand alone will do the work.

Your action list

• If you sell on Depop, check whether any fee, policy, or cross-listing changes have been communicated since the July 30 close, and don’t assume “brand and platform stay the same” means your day-to-day seller experience is fully unaffected going forward.

• If your business touches resale, secondhand, or C2C fashion in any way, look up Vinted’s current US presence and pricing directly. A competitor growing at nearly 50% annually and newly entered into your market is worth a direct look, not just an awareness that it exists.

• If you’re evaluating any acquisition, platform partnership, or category expansion for your own business right now, run the honest version of the question this deal raises: not just whether the category looks attractive today, but whether you’re actually positioned to keep funding the fight for it three to five years out if a well-capitalized competitor shows up in the meantime.

Sources: CNBC, "Vinted's CEO says the US is an 'enormous opportunity' as the $9 billion secondhand marketplace plots its Atlantic crossing" (June 9, 2026); FashionUnited, "Vinted announces 'major' US expansion to address growing demand" (January 22, 2026); Hargreaves Lansdown/Sharecast, "Vinted reports strong revenue growth, profits decline" (April 9, 2026), EcommerceBytes, "eBay to Operate Depop as It Seeks Synergies" (August 2, 2026); EcommerceBytes, "eBay Pays Etsy $1.2 Billion to Expand Its Focus on Fashion" (July 17, 2026); FashionUnited, "eBay completes 1.4 billion dollar Depop acquisition" (July 30, 2026); citybiz, "Etsy Completes $1.4 Billion Sale of Depop to eBay"; StockTitan, "Etsy Completes About $1.4B Depop Sale to eBay," citing Etsy, Inc. press release (July 30, 2026); MarketScale, "eBay closes Depop acquisition and raises 2026 GMV outlook to 12.5% growth as online resale consolidates" (August 14, 2026); Retail Systems, "eBay and Etsy to buy Depop for $1.2bn" (February 2026); Retail Dive, reporting on Etsy CEO Kruti Patel Goyal's comments on workforce reduction (August 6, 2026); CNBC, "Vinted's CEO says the US is an 'enormous opportunity' as the $9 billion secondhand marketplace plots its Atlantic crossing" (June 9, 2026); FashionUnited, "Vinted announces 'major' US expansion to address growing demand" (January 22, 2026).

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually happening

The China Securities Regulatory Commission approved Shein’s Hong Kong listing on July 10, clearing a path after New York and London both fell through. JPMorgan Chase, Goldman Sachs, and Morgan Stanley are joint sponsors. The draft prospectus filed July 26 doesn’t disclose share count, pricing, or the listing date yet, that comes with the investor roadshow and formal bookbuilding, but it does disclose something far more useful to any operator watching this space: real, audited numbers on what the end of duty-free treatment for low-value imports actually did to the business absorbing more of that volume than any other company on earth.

The topline: 2025 revenue grew 8% to $41.8 billion, a sharp deceleration from 20.7% growth the year before. Full-year 2025 net profit fell 38.7%. Then Q1 2026 turned negative entirely, a $99 million net loss against $9.05 billion in barely-grown revenue. Two mechanisms drove that swing, and they’re worth separating clearly because only one of them is really about tariffs. The US ended its de minimis exemption in May 2025, and Shein disclosed that Chinese-origin products it or its marketplace sellers ship to the US now face tariffs ranging from 10% to 87.5% depending on category, a range wide enough that “tariffs went up” understates what actually happened to specific product lines. The second driver was a $328 million non-cash fair-value loss on convertible redeemable preferred shares, an accounting adjustment tied to how those instruments get valued ahead of a public listing, not an operating cost. Strip that one-time charge out and the underlying operating deterioration is still real, but meaningfully smaller than the headline loss makes it look.

Buried in the same filing, disclosed almost in passing relative to the tariff story: Shein’s US business is under investigation by the Federal Trade Commission. The company offered no detail on the nature of the inquiry, said an FTC settlement is possible but the outcome and timing can’t be predicted, and warned investors it can’t rule out the investigation resolving with “significant monetary payments.” An FTC spokesperson didn’t comment. This is the first public disclosure of that investigation existing at all, and it landed inside an IPO document rather than through any prior announcement, which is itself a signal about how these things tend to surface: not proactively, but when a company is legally required to tell investors everything material before asking them for money.

Why real numbers from inside the company matter more than industry estimates

This newsletter has covered the de minimis story from the outside for months, the EU’s €150 threshold ending, the US version ending back in May 2025, the operational scramble that followed for cross-border sellers. All of that coverage necessarily worked from aggregate industry data, carrier volume statistics, survey results, third-party estimates of what the policy shift was doing to the sector broadly. Shein’s filing is different in kind, not just degree. It’s one specific company’s actual, audited income statement, showing precisely how a policy change that this newsletter and dozens of others described in the abstract translated into a $494 million swing from profit to loss in a single company’s US business over twelve months.

That specificity matters for calibrating your own expectations if your business has any meaningful cross-border exposure. Shein is about as extreme a test case as exists, a business built almost entirely around small-parcel, direct-to-consumer shipments from Chinese manufacturing, which is precisely the model de minimis removal was aimed at disrupting. If a company operating at Shein’s scale and sophistication, with resources most DTC brands can’t match, still swung to a loss and saw US revenue drop 14.3% in a single quarter, that’s a real data point about how severe this specific policy mechanism actually is, not a worst-case hypothetical someone constructed for a trade publication.

There’s a second, quieter signal in the filing worth taking seriously: Shein disclosed it has been raising prices in the US since May 2025 specifically to offset a portion of the higher tariff costs, and has been doing so consistently for over a year. A company built on extremely low price points, $5 dresses, $10 jeans, has apparently concluded that absorbing the full tariff cost without any price adjustment wasn’t viable even for a business with Shein’s manufacturing and logistics scale. If the company most associated with rock-bottom pricing is passing tariff costs through to customers rather than eating them entirely, that’s a real signal about where the actual limits of absorption sit, useful information whether you’re setting your own pricing strategy or trying to gauge how much room competitors in adjacent categories actually have.

What this means for how you read the next company disclosure like this

Separate one-time accounting charges from genuine operating deterioration before drawing conclusions from any headline loss number, your own or a competitor’s. Shein’s $99 million loss looks catastrophic as a single figure. Strip out the $328 million non-cash preferred-share valuation charge and the underlying picture, while still genuinely worse than a year earlier, is a different and more specific story about tariff-driven margin compression rather than a business in general collapse. The same discipline applies to your own reporting: a headline number without the one-time items broken out tells you less than the number with them separated.

Treat a company’s pricing response as real information about its actual cost absorption capacity, not just a business decision. Shein raising US prices since May 2025 despite scale and manufacturing advantages most brands can’t replicate is a data point about how much of a 10% to 87.5% tariff range genuinely can’t be absorbed through supply chain efficiency alone. If you’ve been debating whether to pass tariff costs through to your own customers or continue absorbing them, this is a real signal from an operator with far more scale leverage than most, concluding that some pass-through was necessary anyway.

Watch what happens next with the FTC disclosure specifically, since it’s a genuinely open question with real implications beyond Shein itself. An investigation surfacing through a mandatory IPO disclosure rather than a prior announcement is a pattern worth remembering: regulatory scrutiny of a company doesn’t always become visible when it starts, it becomes visible when the company is legally compelled to tell someone. If you compete against or source through platforms with meaningful US regulatory exposure, that exposure may already exist without being publicly known yet.

Your action list

• If your business has cross-border exposure similar in kind, if not in scale, to what Shein disclosed, pull your own landed cost data since your relevant de minimis or tariff change took effect and calculate the actual percentage swing in gross margin, the way Shein’s filing makes visible for its own business. A specific number is more useful for planning than a general sense that costs went up.

• Review your own pricing strategy against the signal in Shein’s disclosure: if a company with Shein’s scale and manufacturing leverage concluded it needed to raise prices rather than fully absorb tariff costs, that’s a reasonable benchmark to weigh against your own instinct to protect price points at the expense of margin.

• If you’re evaluating a competitor, platform, or supplier relationship where regulatory risk is a real but currently invisible variable, treat the Shein FTC disclosure pattern as a reminder that the absence of a known investigation isn’t the same as the absence of one. Build that into how confidently you rely on any single platform or partner’s currently clean public record.

Sources: CNBC, "Fast-fashion retailer Shein's 3-year IPO odyssey may have cost it the 'golden time' to go public" (July 31, 2026); CNBC, "Shein reveals key financials ahead of Hong Kong IPO" (July 26, 2026); Retail Dive, "Shein's US ambitions hindered by FTC investigation, tariffs" (July 28, 2026); Forbes, "Shein Long-Awaited IPO Just Got Harder To Sell As Tariffs Bite" (July 28, 2026); Quartz, "Shein is disclosing an FTC investigation as it readies for Hong Kong IPO" (July 28, 2026); IBTimes UK, "Shein Hit by Tariff Shock as Hong Kong IPO Filing Reveals $99 Million Loss."

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually happening

New York General Business Law Section 396-b took effect June 9, requiring any business that creates a commercial advertisement, and has actual knowledge it contains a synthetic performer, to make a conspicuous disclosure. Amazon notified sellers on July 22 that it was building the compliance mechanism for that law directly into Seller Central: any product image or A+ content containing a photorealistic, fully AI-generated person now needs the keyword “contains-synthetic-performer” written into the file’s XMP metadata, specifically the dc:subject field, using an IPTC-compatible metadata editor, before the file gets uploaded. Once that metadata is present, Amazon adds a shopper-facing disclosure to the listing. Amazon’s own product image guide doesn’t spell out a grace period or a specific penalty tied to the tag itself, but the platform’s general image policy already allows it to pull noncompliant images and suppress a listing from search entirely if it lacks a compliant main image.

The scope is narrower than most of the headlines suggest, and getting the boundary right matters because mistagging in either direction creates real problems. The rule covers only fully synthetic people, generated entirely by AI, with no real person behind them. A real model whose photo got AI-retouched, relit, or had the background swapped still counts as a real person and needs no tag. Characters from movies, TV, games, or other expressive works are exempt. Non-photorealistic illustrations are exempt. But Amazon’s own guidance goes further than the New York statute’s narrow definition in one specific way: a synthetic figure appearing in the background of a lifestyle scene, not the featured model, still requires the tag, even though that background figure isn’t a “performance” in the way the law’s language technically describes.

The part almost every piece of coverage is getting muddled is how this interacts with the other AI transparency law that became operative on exactly the same date this newsletter’s own Article 50 piece covered two weeks ago, and a third law that also just kicked in. California’s AI Transparency Act, SB 942 as amended by AB 853, became operative August 2, the same day as the EU AI Act’s Article 50 transparency rules. But California’s law has a fundamentally different scope than either the EU rule or Amazon’s New York-driven policy: it applies to “covered providers,” meaning companies that build and operate generative AI systems with over one million monthly California users, not to individual sellers or brands using AI tools to generate content. If you’re a DTC brand or Amazon seller using an AI image generator, California’s law creates obligations for the tool you’re using, not directly for you as the user. Amazon’s New York-driven tagging rule is the one that puts the compliance burden on you specifically, as the seller uploading the file.

Why three overlapping laws in three weeks is a real operational problem, not just noise

The instinct is to treat a cluster of regulatory deadlines landing close together as background noise, since no single one of them feels urgent in isolation. That instinct is wrong here for a specific, mechanical reason: these three laws have three different scopes, three different enforcement points, and three different things they actually require, and a seller who reads one summary covering “AI disclosure laws in August 2026” and assumes it’s one unified requirement is going to comply with the wrong thing, or assume they’re covered when they aren’t.

The Amazon rule is the one with the most immediate, concrete operational bite for a typical mid-market e-commerce seller, because it’s enforced at the point of upload by the platform itself, with listing suppression as the real, near-term consequence rather than a theoretical civil penalty. New York’s own statutory penalty is modest on paper, $1,000 for a first violation, $5,000 for subsequent ones, and the law itself exempts platforms that merely publish or distribute the ads, which is precisely why Amazon pushed the compliance obligation down to individual sellers rather than absorbing it as the platform. The real cost isn’t the New York fine. It’s Amazon suppressing your listing for a missing tag, which is a search-visibility and revenue problem regardless of how the underlying state law’s penalty structure is written.

There’s also a scope detail worth taking seriously if you source products or run creative production outside the US. Amazon’s guidance applies at the catalog level, meaning a storefront operated out of Vietnam, China, or anywhere else still has to process its entire image library against the same standard if it’s selling into the US market. This isn’t a US-company problem. It’s a US-marketplace problem, and the seller’s location doesn’t create an exemption any more than it does under the EU AI Act’s Article 50 rules this newsletter covered two weeks ago. Different law, same underlying pattern: the obligation follows the market you’re selling into, not where your business happens to be based.

What to actually check this week, and how not to confuse the three rules

Separate your compliance obligations by which rule actually applies to you before doing anything else. If you’re an Amazon seller, the tagging requirement is yours to implement directly, regardless of your business size or location, and it’s the one with an immediate, platform-enforced consequence. If you build or operate a generative AI tool with over a million California users, California’s law applies to you as the tool provider, which is very likely not your situation if you’re a typical e-commerce brand using an off-the-shelf AI image generator. If EU shoppers can reach your storefront at all, Article 50’s chatbot and content-marking rules apply on top of whichever US-specific rules also touch your business, and none of these three laws substitutes for compliance with the others.

Audit your actual image and A+ content library for fully AI-generated people specifically, not for any AI involvement broadly. The distinction Amazon draws is precise: a real photo that’s been AI-retouched doesn’t need the tag. A fully synthetic person does. If your team or an outside creative agency has been using AI tools loosely across both categories without tracking which images fall into which bucket, that audit itself is the first real piece of work here, before any tagging happens.

If you work with outside creative agencies, freelancers, or contractors producing your product imagery, move the tagging responsibility upstream into your creative brief and delivery contract rather than treating it as a compliance step you’ll handle after the fact. Metadata tagging that has to happen correctly at the point of file creation is a much more reliable process than a post-upload cleanup pass across a catalog nobody’s fully audited. Making synthetic-element identification a delivery requirement, with liability assigned in the contract, is a meaningfully more durable fix than hoping every contractor independently read Amazon’s July 22 announcement.

Don’t assume shopper-facing AI disclosure hurts conversion the way it might feel like it should. Amazon has been vague about exactly what triggers the visible customer-facing label and how prominent it will be, which means the actual conversion impact of a disclosed image versus an undisclosed one isn’t yet established with real Amazon-specific data. Treat this as a genuine open question worth watching rather than a settled cost you’re already pricing in.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

How Levi’s learned the hard way, three years before this rule existed, that framing matters more than the technology itself

In March 2023, Levi Strauss & Co. announced a partnership with Lalaland.ai, an Amsterdam-based AI studio, to generate synthetic models wearing Levi’s products across a range of body types, ages, and skin tones. The stated goal was inclusion: showing shoppers a wider variety of people in Levi’s clothing than a standard photoshoot could economically produce. The backlash was immediate and came from an unexpected direction, not concern about AI imagery itself, but anger that a brand would use synthetic diversity as a substitute for actually paying a diverse group of real models.

The numbers: Levi’s initial announcement drew criticism sharp enough that the company issued a clarifying statement within six days, on March 22, followed by a second statement on March 28 after the backlash intensified further. Shawn Grain Carter, a fashion business management professor at the Fashion Institute of Technology, told NBC News the economics were the real story underneath the diversity framing: “When you have to hire a model, book an agency, have a stylist, do the makeup, feed them on set, all that costs money. Let’s make no mistake about it, Levi’s is doing this because this saves them money.” Levi’s own spokesperson denied cost-saving was the intent. The company walked back the diversity framing specifically, stating in its follow-up that the AI models were not “a means to advance diversity” and would supplement, not replace, real photoshoots.

Why it matters now: Levi’s crisis happened three years before Amazon’s synthetic-performer tagging rule or New York’s disclosure law existed, but it’s the clearest real-world preview available of exactly the dynamic these new rules are built around. The backlash wasn’t primarily about whether AI-generated people are acceptable in commerce. It was about the gap between what a brand says an image represents and what it actually is. A 2026 retrospective on the campaign in CampaignsLive made the sharper point directly: Levi’s use case, e-commerce thumbnails showing product on different body types, was reasonable on its own terms. What broke was the framing, presenting synthetic models as a diversity initiative rather than a production tool. The technology wasn’t the problem. The undisclosed gap between claim and reality was.

How to avoit it:

Treat mandatory disclosure, like Amazon’s new tagging requirement, as a floor, not a ceiling. Levi’s technically complied with nothing that existed in 2023 and still faced a genuine trust crisis, because the backlash was about honesty and framing, not about a missing metadata tag. A disclosed synthetic image used honestly draws far less scrutiny than an undisclosed one framed as something it isn’t.

If you’re using AI-generated people to represent diversity, inclusivity, or any claim about who your brand serves, be direct internally about whether the underlying goal could be served by actually paying a broader range of real people, and be honest publicly about which one you chose and why.

Separate the production-efficiency use case from the brand-story use case before you launch anything. Using AI imagery to generate more product-on-body variations efficiently is a defensible operational choice. Presenting that same choice as evidence of a values commitment is where brands keep getting burned, in 2023 and, per the new tagging rules, potentially now with real platform-level consequences attached.

Expect your customers to notice the gap faster than your own team does. Levi’s backlash moved from announcement to walk-back in under a week. In an environment where disclosure is now a legal requirement on major platforms, not just a reputational risk, that timeline is likely to compress further, not lengthen.

Your action list

• Pull your product image and A+ content library and sort it specifically into fully AI-generated people versus everything else, real people with any degree of AI editing, non-photorealistic content, or no people at all. Only the first category needs the tag.

• If any of your creative production runs through an outside agency or freelance contractor, update your brief and delivery requirements this week to make synthetic-element tagging part of what they deliver, rather than something you catch after the file is already uploaded.

• Confirm which of the three current AI disclosure regimes, Amazon’s New York-driven seller rule, California’s covered-provider law, or the EU’s Article 50, actually applies to your specific business, since conflating them is the most common mistake showing up in coverage of this right now, and each one requires a genuinely different compliance action.

Sources: CNBC, "Amazon makes sellers label AI-generated people in images after NY law" (July 23, 2026); MLQ News, "Amazon Requires Sellers to Tag AI-Generated People in Product Images After New York Law" (July 24, 2026); Forbes, "Amazon Requires Sellers To Label AI-Generated People In Listing Images" (July 25, 2026); Five Star Commerce, "New Amazon AI Image Rule: How to Tag AI-Generated People in Product Images"; Goat Consulting, "Amazon's AI-Generated Image Rule: What Sellers Must Do"; Masonry Blog, "Amazon AI Product Image Rules for Sellers (2026)"; Morgan Lewis, "New California AI Disclosure Rules Become Operative" (August 2026); Mayer Brown, "New Obligations Under the California AI Transparency Act and Companion Chatbot Law Add to the Compliance List"; AI Laws By State, "California AI Transparency Act (SB 942): 2026 Compliance Guide.", NBC News, "AI models Levi's controversy backlash" (2023); Rangefinder, "After backlash, Levi's says AI-generated models 'to increase diversity' won't replace real shoots" (2023); Silicon Canals, "Levi Strauss clarifies its use of Lalaland's AI models after online backlash"; Journal of Contingencies and Crisis Management, "Levi's and Lalaland.ai collaboration crisis," Maiolo, 2024; CampaignsLive, "The Levi's + Lalaland.ai Controversy: What It Got Wrong" (March 2026))

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually happening

Sidekick launched in January 2025 as a chat interface built directly into the Shopify admin, reachable through a purple glasses icon in the top bar. It’s powered by Anthropic’s Claude, running on Google Cloud’s Vertex AI, and its defining feature is direct access to a merchant’s actual store data, products, orders, settings, analytics, rather than operating as a generic assistant bolted onto the platform from outside. A merchant can ask it why sales dropped last week, have it draft a discount campaign, generate a performance report, or build a custom app, and Sidekick executes inside the admin with the merchant’s approval rather than sending them off to read documentation or hunt through settings menus.

The adoption numbers Shopify disclosed this quarter are large enough to signal something structural rather than a niche feature a small subset of merchants happen to use. Daily active merchants using Sidekick grew 3.6x year over year. Daily sessions grew 4.8x. The assistant handled roughly 34 million conversations in the quarter alone and was used to build about 36,000 custom apps, up from 12,000 the quarter before. During the 2025 BFCM period, more than 750,000 shops used Sidekick for the first time in a single quarter, according to Shopify’s own published case coverage, using it for last-minute copy rewrites and campaign prep under real time pressure heading into the platform’s biggest sales weekend.

The specific onboarding statistic is the one worth sitting with longest, since it’s the rare disclosed number that ties an AI feature directly to a business outcome rather than just usage volume. An 8% lift in merchants reaching five orders within 15 days isn’t a vanity metric about how many people opened a chat window. Five orders in the first two weeks is a real, meaningful early-traction threshold, the kind of number that correlates with whether a new store finds real footing or stalls out in its first month. Shopify tying Sidekick usage specifically to that threshold, rather than to a softer metric like session length or feature adoption, suggests the company itself is treating early order velocity as the outcome that actually matters, and found that merchants using the assistant during setup were meaningfully more likely to hit it.

Why this matters beyond Shopify’s own stock price

This newsletter has spent a lot of space this year on the discovery-layer version of AI in commerce, agentic shopping protocols, AI Overviews reshaping search traffic, AI-referred visitors converting better than any other channel. Sidekick is a different category of AI entirely, and it’s worth distinguishing the two clearly. Discovery-layer AI changes how customers find and evaluate your store. Sidekick is operations-layer AI, aimed entirely at the merchant side of the relationship, helping the person running the store make faster, better-informed decisions about their own operation. Shopify’s own framing on the earnings call reinforced that split: catalog-powered AI search, the customer-facing discovery layer, converted at roughly twice the rate of scraped-data search, while agentic traffic and orders to Shopify stores tripled year over year. Both are real and both matter, but they’re solving different problems, one on the demand side, one on the operator’s own side of the admin panel.

For a new or growing store specifically, the operations-layer version may matter more in the near term, precisely because it addresses a bottleneck that has nothing to do with traffic or demand. A huge share of stores that fail to gain traction in their first month aren’t failing because customers can’t find them. They’re failing because the founder or small team running the store is drowning in setup tasks, product descriptions, discount configuration, basic analytics interpretation, that eat the time and attention that should be going toward the handful of decisions that actually move the needle. An assistant embedded directly in the admin, with real access to the store’s own data, cutting through that setup friction, is a plausible, mechanistically sound explanation for why order velocity in the first 15 days would move at all. It’s not solving demand generation. It’s solving execution speed during the exact window when execution speed determines whether a store lives past its first quarter.

There’s a second, quieter implication in the adoption numbers themselves. A 3.6x year-over-year jump in daily active merchants using a free, built-in tool suggests this has crossed from early-adopter usage into a genuinely mainstream workflow for a meaningful share of Shopify’s merchant base. If you’re running a Shopify store and haven’t used Sidekick beyond a curious first click, you’re increasingly the exception rather than the norm among active merchants on the platform, and the specific onboarding data suggests that gap may be showing up in your own early-stage metrics without an obvious cause you’d otherwise trace back to it.

What to actually check if you’re running or launching a Shopify store

Check whether you’re actually using Sidekick for the setup-phase tasks it’s disclosed to help with most, not just the flashy content-generation use cases. The 8% lift showed up specifically during onboarding, which suggests the highest-value early use isn’t necessarily asking it to write a blog post. It’s the more mundane setup work: catalog structuring, discount configuration, initial analytics interpretation, the friction-heavy tasks that eat founder time in a new store’s first weeks without directly generating a sale.

If you’re past the onboarding window already, don’t assume the tool has nothing left to offer. Shopify’s disclosed use cases extend well beyond first-15-day setup: ShopifyQL query writing for performance and payments data, customer and company record creation through natural language rather than manual form-filling, and ongoing analytics interpretation that would otherwise require pulling and reading reports manually. If your current workflow still involves manually digging through admin reports to answer basic performance questions, that’s a task Sidekick is specifically built to shortcut.

Check what’s actually driving your own new-product or new-campaign launches right now, and whether setup friction is quietly costing you time you’re not accounting for. If launching a new product or promotion on your store currently takes a half day of manual configuration, description writing, and settings adjustment, that’s exactly the category of task the disclosed usage data suggests this tool meaningfully accelerates, and it’s worth testing directly against your own actual launch timeline rather than assuming the improvement is marginal.

Don’t confuse this with the discovery-layer AI story this newsletter has covered extensively elsewhere. Using Sidekick well doesn’t make your store more visible to AI shopping agents or improve your catalog’s machine readability for ChatGPT or Gemini. It’s a separate problem, solved by separate infrastructure, structured product data, protocol support, robots.txt configuration, covered in earlier issues. Sidekick speeds up how you run the store. It doesn’t determine whether AI systems can find and recommend it.

Your action list

• If you haven’t used Sidekick beyond an initial glance, pick one real, current task, drafting a discount campaign, interpreting last month’s conversion data by traffic source, auditing a product page for clarity, and run it through the assistant this week as an actual test rather than a curiosity click.

• If you’re currently onboarding a new store or launching a new product line, deliberately route the setup-heavy tasks, catalog structuring, initial content, discount configuration, through Sidekick rather than doing them manually, and track whether your own time-to-first-orders shortens against your prior launches.

• Audit your current admin workflow for any recurring task that involves manually pulling reports or digging through settings to answer a question you could instead just ask directly. If that list is longer than a couple of items, that’s your actual starting point for where this tool is most likely to save real time, not the content-generation use cases that tend to get the most attention.

Sources: Shopify Inc., "Shopify Announces Second Quarter 2026 Financial Results," Form 8-K filing (August 5, 2026); TipRanks, "Shopify Earnings Call Signals Durable High-Growth Runway" (August 2026); TipRanks, "Shopify Inc. (SHOP) Q2 2026 Earnings Report"; Investing.com, "Shopify Q2 2026 slides: AI commerce drives 32% GMV growth"; Yahoo Finance/GuruFocus, "Shopify Inc (SHOP) (Q2 2026) Earnings Call Highlights"; InsiderFinance, "Shopify Q2 2026 Earnings Beat Estimates" (August 2026); Zacks Investment Research via Yahoo Finance, "SHOP Q2 Earnings Beat Estimates, Revenues Ride on Strong GMV"; Shopify, "Meet the AI tool that's helping independent brands compete with retail giants this holiday season" (November 5, 2025); Google Cloud, "How Shopify scales AI-powered retail to empower millions of merchants with Claude on Vertex AI."

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What actually happened

The client was spending most of their acquisition budget on Meta, running a mix of prospecting and retargeting, with blended ROAS as the north star metric everyone in the business, including me, was checking every Monday. It looked healthy. It had looked healthy for months. Nobody had a reason to dig deeper, because the number on the dashboard was doing exactly what a good number is supposed to do: sit there quietly and not cause alarm.

The thing that made me finally check was almost an accident. I was building a cohort view for a different question entirely, trying to understand repeat purchase rate by acquisition channel, and I needed to separate new customers from returning ones inside the Meta data itself. Meta doesn’t make that easy by default. You have to actually dig for it, pull the new-versus-existing customer breakdown, and cross-reference it against your own backend order data rather than trust what the ad platform hands you.

Once I did that, the 4.2x number split into two very different stories. A meaningful share of what Meta was reporting as ROAS-driving conversions were customers who had already bought from the brand before, some of them multiple times, being served retargeting ads and getting credit for a purchase they were extremely likely to make anyway. Strip those out and look only at genuinely new customers, and the real acquisition ROAS was closer to 1.6x. Still positive. Nowhere near what the dashboard had been telling everyone for months.

Why I missed it, and why that’s the actually useful part

I want to be honest about the reason I missed it, because the reason is more useful than the mistake itself. I know how attribution works. I’ve written about the exact mechanism, Meta’s modeled conversions, the post-iOS 14 measurement gap, the tendency for platforms to over-credit retargeting, more than once. Knowing the mechanism in the abstract and actually checking for it on a live account are two different disciplines, and I let the first one substitute for the second because the top-line number looked fine and there was always something more urgent to look at that week.

That’s the trap. A blended ROAS number that looks healthy removes the pressure to dig into it, and the digging is exactly the work that would have caught the problem three months earlier. The client wasn’t losing money. The business wasn’t in trouble. But a meaningful share of the ad budget was going toward reinforcing purchases that were going to happen anyway, money that could have gone toward actually finding new customers instead, and nobody would have found that gap by staring at the same dashboard every Monday, because the dashboard was never built to show it.

What I changed after that

The fix wasn’t complicated once the problem was visible. We built a standing weekly view, separate from the main ROAS dashboard, that split spend and conversions by new versus returning customer specifically, pulled from backend order data rather than Meta’s own attribution. Retargeting spend got its own line item with its own target, evaluated on its own terms rather than blended into a number that made prospecting look better than it was. Within a month, we’d reallocated a real chunk of budget away from retargeting customers who barely needed the nudge and toward prospecting that was actually finding people the brand hadn’t reached before.

The number that mattered wasn’t a new metric. It’s one this newsletter has flagged before in different contexts: are you measuring blended performance or are you measuring the specific thing you’re actually trying to buy. A blended ROAS answers a vague question well. It answers the actual question, is this budget finding new customers or just harvesting existing ones, terribly, and it took me three months on a live account to remember that the hard way instead of just knowing it as a fact I could recite.

What to check in your own account this week

Pull your new-versus-returning customer split for your top-spending Meta campaigns over the last 90 days, cross-referenced against your own backend order data, not just Meta’s own reporting. If a meaningful share of what’s driving your blended ROAS is retargeting spend against customers who were already going to buy, that’s not a wasted line item exactly, but it’s being counted as acquisition performance when it isn’t, and that distinction changes how you should be evaluating and budgeting it.

Separate your prospecting and retargeting spend into genuinely separate views with separate targets, not just separate campaigns feeding into the same blended number. A retargeting campaign converting at a high rate against existing customers and a prospecting campaign converting at a lower rate against genuinely new ones will always make the blended number look better than either one actually deserves credit for on its own.

Don’t assume knowing how attribution works protects you from the exact problem attribution creates. I’ve written the explainer on this mechanism more than once, and I still let a live account run for three months on a number I should have questioned from week one. The fix isn’t more knowledge. It’s a standing habit of checking the specific number, not the blended one, on some kind of actual schedule rather than waiting for a reason to look.

Your action list

• Pull new-versus-returning customer data for your highest-spend ad campaigns over the last 90 days, sourced from your own backend orders, not the platform’s own attribution, and see how much of your reported ROAS is coming from customers who would very likely have bought anyway.

• If retargeting and prospecting are currently blended into one reported number anywhere in your regular reporting, split them this week into separate views with separate targets, even if the underlying campaigns stay structured the way they already are.

• Set a recurring calendar reminder, monthly is enough, to actually run this specific check rather than trusting that a healthy-looking top-line number means there’s nothing underneath it worth questioning. The dashboard that looks fine is exactly the one that doesn’t get a second look, which is precisely how something like this sits unnoticed for months.

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually happening

Article 50 of the EU AI Act becomes enforceable today, and it covers three distinct obligations that apply regardless of how big your business is or whether you’re even based in the EU, only whether EU users interact with your systems. Chatbots and other interactive AI systems have to tell users they’re talking to AI, not a person, and this specific obligation applies from today without exception, it was not touched by the delays affecting other parts of the Act. AI-generated or manipulated content, image, video, audio, or text, intended for public distribution has to carry a machine-readable mark showing it was artificially generated, and where possible, that mark needs to be perceptible to the person seeing it too. Deepfakes specifically, AI content depicting a real, identifiable person doing or saying something they didn’t, get a stricter labeling standard on top of the general rule.

One distinction worth getting exactly right, since several sources blur it: the content marking and labeling requirement carries a transitional period until December 2, 2026 specifically for AI systems already on the market before today. The chatbot disclosure duty does not get that same grace period. If you’re already running an AI chatbot, you needed disclosure live as of today. If you’re already publishing AI-generated content without a marking system, you have a narrower window to fix that specific piece before the December date, though building the habit now rather than waiting is the safer read given how new this enforcement regime is.

The Commission’s own July 31 announcement is direct about the scope: this isn’t limited to systems classified as high-risk, and it isn’t limited to companies headquartered in Europe. It applies to any AI system EU users interact with, which for an e-commerce operator means the customer service chatbot on your storefront, an AI product-description generator whose output goes live on a product page an EU customer can reach, or an AI image tool used to create marketing content that ends up in front of EU shoppers. The trigger is the EU user’s exposure, not your company’s location or size.

The penalty structure explains why this deadline is getting real attention rather than being treated as background regulatory noise. Violations of prohibited AI practices carry fines up to €35 million or 7% of global annual turnover, whichever is higher. Article 50 transparency violations specifically, along with high-risk system violations, carry fines up to €15 million or 3% of global turnover. For a mid-market e-commerce brand, even the lower tier is a genuinely serious number relative to typical DTC revenue, not a fine you’d treat as a cost of doing business the way some regulatory penalties get absorbed.

Why this is easy to miss even if you’re paying attention to trade and tariff news

Most of the regulatory coverage e-commerce operators have been tracking this year has been about tariffs, and for good reason given how much this newsletter alone has covered on that front. Article 50 sits in a completely different category, an operational compliance requirement with an actual court-enforceable deadline, and it’s landed with far less warning than any of the tariff changes did. Tariff rate changes get covered by every trade publication and customs broker for weeks in advance. A chatbot disclosure requirement doesn’t generate the same volume of pre-deadline coverage, even though the penalty exposure is arguably more severe for a brand that’s simply never turned its attention to AI regulation at all.

There’s also a genuine source of confusion baked into how this rule has been reported, worth clearing up directly. Multiple outlets covering the EU AI Act this year have focused heavily on the high-risk system obligations, which did get delayed through the Digital Omnibus process, sixteen months for the standalone high-risk category, twelve months for embedded high-risk systems. A brand that read that delay news and concluded the entire AI Act got pushed back is now walking into an enforceable deadline it thinks doesn’t apply yet. The delay was real, but it applied to a different, narrower set of obligations than the ones taking effect today. Article 50 was never part of that deferral.

What actually counts as compliant, and what doesn’t

The chatbot disclosure requirement doesn’t require a complex system. It requires that a user interacting with an AI-driven chat interface be told, in some clear way at the point of interaction, that they’re talking to AI rather than a human agent. A simple, visible statement at the start of the chat, something identifying the assistant as AI-powered, meets the core requirement. What doesn’t meet it is a chatbot designed or presented in a way that could reasonably lead a user to believe they’re talking to a person, which is a design and copy question as much as a technical one. If your chat widget uses a human name, a photo of a person, and conversational copy with no AI disclosure anywhere in the interface, that’s the exact pattern this rule targets.

The content labeling requirement is broader and easier to overlook. Any AI-generated or AI-altered image, video, audio, or text intended for public distribution needs a machine-readable mark, and where feasible, a mark a human can actually perceive too. That covers more of a typical e-commerce operation than the chatbot rule alone. AI-generated product photography, AI-written product descriptions or marketing copy, AI-edited lifestyle imagery, any of it, if it’s public-facing and reaches EU users, falls under this requirement. The deepfake-specific rule is stricter still: AI content depicting a real, identifiable person requires disclosure regardless of context, which matters directly for any brand using AI to generate spokesperson content, testimonial-style video, or influencer-style imagery.

One detail worth flagging honestly rather than glossing over: enforcement mechanics and national-level guidance are still forming in real time. The Commission’s own announcement points to national authorities as co-enforcers alongside the EU AI Office, and how aggressively national regulators pursue smaller e-commerce operators in the first weeks and months after a deadline like this is genuinely uncertain. The rule is enforceable starting today. How quickly and how broadly that enforcement actually reaches mid-market DTC brands, as opposed to large platforms and GPAI providers, is not yet established by precedent.

What to actually check this week

Audit every AI-facing touchpoint on your storefront that an EU customer could reach, not just your primary market. If you sell into the EU at all, even as a secondary market, your chatbot, your AI content, and your AI-generated media are all in scope, regardless of whether your business is headquartered in Europe. Company location doesn’t determine applicability. User location does.

Check your chatbot’s actual point-of-interaction disclosure, not whether one exists somewhere on your site in a footer or terms page. The requirement is disclosure at the point of interaction, meaning the user needs to know they’re talking to AI at or near the moment the conversation starts, not buried three clicks away in a privacy policy.

Inventory your AI-generated public content specifically, since this is the requirement most operators are least likely to have already addressed. If you’ve used AI tools for product photography, marketing copy, or promotional video and haven’t implemented any labeling, that’s a real gap between where your operation currently stands and where Article 50 requires it to be as of today.

If you use AI to generate any content resembling a real, identifiable person, spokesperson-style video, AI-generated testimonials, synthetic influencer content, treat that as the highest-priority item on this list. The deepfake-specific disclosure standard is stricter than the general labeling rule, and it’s the category most likely to draw early enforcement attention given the stated goal of reducing deception and manipulation.

Your action list

• Pull a full list of every AI-driven touchpoint on your storefront and marketing channels that an EU user could encounter, chatbots, AI-generated images, AI-written copy, and confirm each one against Article 50’s three specific requirements: AI disclosure for interactive systems, machine-readable marking for generated content, and the stricter deepfake standard for any content depicting a real person.

• If your chatbot doesn’t currently disclose it’s AI at the start of the interaction, fix that this week. It’s a low-effort, high-priority change given today’s enforcement date, not a project to schedule for next quarter.

• Check whether your business, or any AI content vendor you use, appears on the Commission’s published list of organizations that have signed the Code of Practice on transparency of AI-generated content. If you’re using a third-party AI tool for product content or customer service, ask that vendor directly what compliance support they’re providing for Article 50 rather than assuming it’s handled on their end.

Sources: European Commission, Directorate-General for Communications Networks, Content and Technology, "Commission starts enforcing AI Act rules and new transparency requirements on 2 August" (July 31, 2026); European Commission, full press release, ec.europa.eu/commission/presscorner (IP_26_1714); Qualimero, "EU AI Act Chatbot Compliance for E-Commerce" (June 2026); Need to Know AI, "EU AI Act Chatbot Disclosure Rules" (July 2026); Connect On Tech (Baker McKenzie), "New EU guidance on AI transparency: what should companies be doing from 2 August 2026"; Legiscope, "EU AI Act: Practical Compliance Guide for 2026" (July 2026); TechJack Solutions, "EU AI Act Article 50: August 2 Compliance Requirements" (July 2026).

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually happening

CEO Steve Huffman opened the earnings call with a specific framing worth sitting with: as the internet becomes more automated, the value of authentic human conversation rises, and Reddit’s business lives with direct, repeat users rather than what he called drive-by search traffic. That’s not just a mission statement. It’s a direct answer to the question investors actually asked on the call, whether Reddit’s growth is exposed to Google’s ongoing search volatility and AI Overviews eating into referral traffic the way this newsletter has covered happening across the rest of the web. Huffman’s answer was that Reddit’s ad business doesn’t depend on that traffic. It depends on people who open the Reddit app directly, and he said those direct users are worth multiples more than anyone arriving via a search click.

The number behind that claim is real and specific: Reddit crossed $1 million in revenue per employee this quarter, alongside a 43% adjusted EBITDA margin, both framed by leadership as long-chased milestones finally hit. Daily active uniques grew 18% year over year to 126.8 million. This is not a story about a platform struggling to find its footing. It’s a platform posting numbers most public companies would call a career quarter.

The stock drop traces to something narrower and more specific: guidance and buyback pace, not the quarter itself. Reddit had only repurchased $5 million against a $1 billion buyback authorization through Q1, a detail flagged as a live concern heading into this report, and a reported dispute over the terms of Reddit’s data-licensing relationship with Google added uncertainty analysts were already pricing in before results even came out. None of that changes what the advertising business itself is doing. It changes how confident investors are that the growth rate holds through the back half of the year against tougher comparisons.

Why this matters for an e-commerce operator, not just Reddit shareholders

Strip away the stock reaction and look at what Reddit’s own advertiser-facing product data shows, since that’s the part that actually determines whether the platform belongs in your media mix. Reddit Max, the platform’s automated bidding and targeting system, alpha-tested across more than 600 advertisers, delivered 17% lower cost per action and 25% to 27% more conversions than standard campaign setups, depending on which quarter’s disclosure you check. Dynamic Product Ads, the format built specifically for retail and e-commerce advertisers, produced more than 90% higher return on ad spend compared to advertisers’ prior campaigns. TransUnion’s independent meta-analysis of EMEA retail advertisers running on Reddit between January 2023 and December 2025 found a 7x average return on ad spend, third-party research Reddit is using specifically to make the commercial case to skeptical performance marketers.

The Shopify integration is the detail worth understanding mechanically, since it’s what actually lowers the barrier for a mid-market DTC brand to test the platform at all. Reddit’s native Shopify integration moved from limited alpha testing to full global availability on May 27, 2026, at no cost through the Shopify App Store. Once connected, a Shopify merchant’s product catalog, pricing, imagery, descriptions, and real-time inventory data syncs directly into Reddit’s ad system, and purchase attribution runs through Reddit’s own pixel without custom development work. That’s a meaningfully lower setup cost than what building a Reddit ads presence required even a year ago, when catalog syncing and pixel implementation were manual, engineering-dependent tasks that priced most smaller brands out of testing the platform seriously.

There’s a specific consumer behavior shift underneath these numbers worth taking seriously rather than treating as marketing copy. Calvin Klein’s global merchant Clay Lute said publicly in January that Gen Z doesn’t trust ads, they trust Reddit, adding that he personally doesn’t make a purchase without checking Reddit first. That’s not a Reddit talking point, it’s an operator at a major legacy brand describing his own actual purchase behavior. High-intent shopping conversations on the platform, threads where people are actively asking for product recommendations or validating a purchase decision, grew 40% year over year according to the company’s own Q2 disclosure. That’s a different kind of intent signal than a scroll-stopping video ad competing for attention on a feed built around entertainment. A shopper reading a Reddit thread asking “is this brand actually good” is closer to bottom-of-funnel than someone watching a TikTok.

What to actually check before testing Reddit as an ad channel

Check whether your product category has genuine, organic conversation happening on Reddit already, before spending a dollar on ads there. Search your own brand name and your closest competitors on the platform directly. Categories with active subreddits built around genuine product comparison and recommendation, skincare, tech, outdoor gear, supplements, personal finance, tend to be where Reddit’s high-intent conversation dynamic is real. A category with no organic Reddit presence at all is a weaker candidate for the platform’s core advantage, regardless of how good the ad tooling has gotten.

If you’re already running Shopify, the actual technical lift to test this is now genuinely low, which changes the cost-benefit calculation of at least running a controlled trial. Since catalog sync and pixel setup are largely automated through the native integration, the real cost of testing Reddit for 30 to 60 days is ad spend and someone’s time to monitor it, not a development sprint. That’s a different decision than it would have been a year ago when the integration required manual work most small teams didn’t have bandwidth for.

Treat Reddit Max as the entry point rather than building manual campaigns from scratch. The 17% CPA reduction and 25%-plus conversion lift reported across alpha testers came specifically from the automated system, not from manually configured campaigns. Given how new the platform still is for most performance marketers, starting with the automated tooling reduces the learning curve you’d otherwise pay for in wasted early spend while you figure out targeting and bidding manually.

Don’t expect Reddit to replace Meta or Google in your mix. Every credible read on this platform, including Reddit’s own leadership, positions it as complementary to search and social rather than a replacement for either. The realistic test is whether it adds incremental, profitable volume on top of what you’re already running, not whether it becomes your primary channel.

Your action list

• Search your brand and your top two competitors directly on Reddit. If there’s real, active conversation, threads asking for recommendations, comparison discussions, reviews, that’s a genuine signal worth testing against. If there’s nothing, that’s useful information too before you spend anything.

• If you’re on Shopify, check whether the native Reddit integration is already available in your app store region, and estimate the real setup time. With catalog sync and pixel automated, this is likely a same-day test to configure, not a multi-week project.

• If you decide to test, start with Reddit Max and Dynamic Product Ads specifically, not manual campaign builds, and run it for a full 30 to 60 days before judging results. The reported performance gains are tied to the automated system learning your account, not to campaigns that are immediately optimized on day one.

Sources: Reddit, Inc., “Reddit Reports Second Quarter 2026 Results,” Form 8-K filing (July 30, 2026); The Motley Fool, “Reddit (RDDT) Q2 2026 Earnings Call Transcript” (July 30, 2026); Investing.com, “Earnings call transcript: Reddit tops revenue forecast in Q2 2026, shares fall 12.5%” (July 30, 2026); CNBC, “Reddit (RDDT) Q2 2026 earnings report” (July 30, 2026); 24/7 Wall St., “Live: Will Reddit’s Q2 Earnings Tonight Send Shares Higher?” (July 30, 2026); PPC Land, “Reddit makes Shopify integration global as EMEA retail data shows 7x ROAS” (May 27, 2026); Proactive Investors, “Reddit integration with Shopify seen driving ad growth, adoption” (May 29, 2026); SocialDay, “Reddit’s Shopify integration goes global: why high-intent shopping conversations now convert” (June 2026); Welfare Capital Research, “Reddit Q1 2026 Earnings Update” (May 4, 2026).

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually happening

The mechanism is straightforward once you see it. A brand sets a trigger keyword, something like “LINK” or “ACCESS,” and publishes it in a caption or story. When a follower comments that word, Klaviyo automatically sends them a DM inviting them to opt into email, SMS, or WhatsApp. One tap, and that follower stops being an anonymous Instagram engagement and becomes a subscriber profile sitting inside the same Klaviyo record that already tracks that person’s browsing history, past orders, and every other touchpoint with your brand. The matching is fuzzy on purpose, typos, emojis, capitalization variants all still trigger the reply, which matters because the entire point is removing friction at the exact second someone’s attention is highest.

What makes this a genuine product shift rather than a minor feature update is where the data actually lands. Instagram DMs, comments, mentions, tags, and posts now write directly into Klaviyo’s unified customer profile, the same one that already holds order history, browsing behavior, and email and SMS engagement. Segments built from that social behavior, Klaviyo’s own example is “commenters who haven’t bought,” sync back out to Meta, Google, TikTok, and Pinterest for lookalike targeting, retargeting, or suppressing existing customers from cold acquisition spend. That’s a closed loop that didn’t exist cleanly before: social engagement becomes CRM data becomes better-targeted ad spend, inside one platform, without a brand needing to stitch together three separate tools to make it work.

The free tier matters here too. Social Auto-replies itself is free on every Klaviyo account. The deeper capture and event-tracking layers sit behind a paid tier, but the core mechanism, comment to DM to subscriber, costs nothing beyond having a Klaviyo account already connected to Shopify, BigCommerce, WooCommerce, or a custom store.

Why this matters beyond one product feature

This newsletter has spent a lot of space on one structural problem: acquisition costs keep climbing while the traffic driving that spend gets harder to attribute cleanly. Average CAC has risen roughly 60% over five years, and the brands still treating paid social as the center of their acquisition model are paying 2026 prices for a channel that stopped delivering 2019 economics years ago. Social Auto-replies attacks a different piece of that same problem. It doesn’t lower your CPM. It converts attention you’re already generating organically, comments, DMs, story replies, engagement your content earned for free, into an owned asset instead of a metric that disappears the moment the algorithm stops showing your post to anyone.

The distinction worth sitting with is rented versus owned. A follower is rented. Instagram controls whether that follower ever sees your content again, and a platform algorithm change can sever that relationship overnight with no warning and no recourse. A subscriber on your own email and SMS list is owned. You control the send, the frequency, and the message, and no platform policy shift can take that list away from you. Every brand with a social following has been sitting on a pool of rented attention that converts into owned relationships only when someone manually clicks a link in a bio, fills out a form, or remembers to visit the site later. Most of that attention just evaporates. Social Auto-replies is a bet that capturing consent at the exact moment of highest engagement, the second someone comments because they actually want something, converts at a rate that manual link-in-bio funnels never came close to matching.

The Evereve and Kulani Kinis numbers back that bet up with real data rather than just product marketing language. A $3.39 cost per subscriber is not a number paid acquisition has delivered in any DTC category this newsletter has covered in the past year. Even accounting for the fact that these are Klaviyo’s own selected case studies, not an independent audit, the mechanism itself, capturing consent at peak engagement rather than hoping for a later click, is a real and specific reason to expect lower acquisition cost than a cold paid ad, which has to first earn attention before it can ever convert anything.

What to actually check before deciding whether this is worth setting up

Check whether your brand already generates comment volume worth converting. This tool doesn’t create engagement, it captures engagement that’s already happening. If your Instagram posts routinely draw genuine comments, not just emoji reactions, you have raw material sitting there right now that’s converting into nothing but vanity metrics. If your comment volume is thin, fix the content problem first, since a keyword campaign on a post with twelve comments won’t move a subscriber count meaningfully regardless of how well the mechanism works.

Check what you’re actually offering as the reason to opt in. Evereve’s result was tied to a giveaway. Kulani Kinis ran a keyword campaign, likely tied to early access or a specific offer. The mechanism handles the capture, but the offer still has to be worth a follower giving up their contact information. A generic “comment LINK for our newsletter” without a real incentive attached will convert at a fraction of what a genuine giveaway or early-access offer does.

Check whether your Klaviyo account is already connected to your storefront platform, since that unified profile is the entire value of this feature. A subscriber captured through Social Auto-replies who lands in a disconnected list, separate from your order and browsing data, is just a slightly more efficient version of the manual link-in-bio funnel you already have. The value compounds specifically because the new subscriber immediately enters the same profile that already knows what they’ve bought, browsed, or abandoned.

Check your current organic-to-paid ratio in your acquisition mix. If nearly all of your customer acquisition budget currently runs through paid social, and organic social exists mainly as a brand-awareness afterthought, that’s worth revisiting given this specific capability. A channel you’ve been treating as unmeasurable exposure just gained a direct, trackable path to becoming a measurable, owned acquisition channel.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

How LSKD turned SMS into its top-performing channel by treating it as community, not a discount blast

LSKD, the Australian activewear brand founded by Jason Daniel in 2007, built its SMS program around a different premise than most DTC brands use for texting. Daniel told Retail Dive that most brands underestimate how powerful a text actually is, and that with everyone already on their phones, SMS is a way to actively engage a community rather than just push a promotion. That framing, engagement over discounting, is what turned SMS into LSKD’s top-performing communication channel.

The numbers: LSKD has scaled to more than $200 million in revenue with over 700 employees and 31 retail stores as of 2026, on track for 41 locations by year end, growing from a brand Daniel started at 16 selling t-shirts out of his mother’s bedroom. SMS became the brand’s top-performing channel specifically through welcome flows, abandoned cart, browse abandonment, and loyalty points reminders, the full lifecycle of a customer relationship rather than a single promotional blast type.

Why it worked: LSKD didn’t treat SMS as a cheaper version of email blasting discount codes. It built the channel around the same community identity that defines the rest of the brand, one built explicitly around sport, fitness, and adventure, with SMS functioning as a direct line into that community rather than an interruption to it. The loyalty points reminder flow specifically ties owned messaging to an existing relationship mechanic, rather than treating every text as a cold sales attempt. That distinction, using an owned channel to reinforce a relationship a customer already has with the brand instead of using it purely to manufacture urgency, is what separated LSKD’s SMS performance from a generic blast strategy.

How to replicate it:

Before writing your first SMS flow, decide what the channel is actually for. LSKD treated it as a community touchpoint, not a discount delivery mechanism, and structured welcome, cart recovery, and loyalty messaging around that identity consistently.

Tie your owned-channel messaging to a real relationship mechanic where you have one, a loyalty program, a rewards balance, a community event, rather than defaulting every message to a percentage-off offer.

If you’re just now converting rented social engagement into an owned list through a tool like Klaviyo’s Social Auto-replies, treat what LSKD did with SMS as the model for the next step. Capturing the subscriber is only the first move. What that channel becomes, discount machine or genuine community touchpoint, determines whether it performs like a top channel or gets ignored like most promotional texts do.

Your action list

• Pull your last 90 days of Instagram comment volume across your top-performing posts. If there’s real, consistent comment activity you’ve never converted into anything beyond a like count, that’s your first candidate for a keyword campaign.

• If you’re already on Klaviyo, check whether Social Marketing is connected to your Instagram Business account, and confirm your storefront integration is live so any new subscriber lands inside the same profile as your existing customer data, not a disconnected list.

• Design one specific offer, not a generic newsletter signup, tied to a single trigger keyword, and test it on your next high-engagement post before rolling it out broadly. A giveaway, an early-access window, or a genuine discount gives people an actual reason to comment and opt in, the way Evereve’s and Kulani Kinis’s results suggest is doing the real work behind those numbers.

Sources: Retail Dive, "How fashion DTC brands engage with their audiences via SMS," featuring comments from Jason Daniel, Founder and CEO of LSKD; company scale confirmed via Fitt Insider Podcast and Style Magazines, 2026, Klaviyo, "Social List Growth: Turn Social Media Followers Into Subscribers" (klaviyo.com); Klaviyo, "Instagram CRM: Convert Comments to Subscribers" (klaviyo.com); Klaviyo Help Center, "How to convert Instagram followers into subscribers with Social Auto-replies"; Digital Applied, "Klaviyo Social Marketing GA: Instagram DMs Meet Your CRM" (July 10, 2026); Threadpoint, "Klaviyo Social Marketing: What It Means for E-Commerce Brands" (2026); Practical Ecommerce, "New Ecommerce Tools: July 15, 2026"; Retail Dive, "How fashion DTC brands engage with their audiences via SMS."

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually in the filing

Reformation’s 2025 gross margin came in at 60.2%, down 360 basis points year over year, and the company attributes that decline directly to IEEPA tariffs in its own filing language. That’s a real, quantified, publicly disclosed number for exactly the kind of tariff cost this newsletter has covered from the policy side for months, now visible from the inside of an actual brand’s income statement rather than industry-wide estimates. Adjusted EBITDA landed at $45 million, 8.9% of net revenue, a healthy margin by DTC standards but one built on a business absorbing real cost pressure it couldn’t fully pass through to customers.

The first quarter of 2026 tells a sharper version of the same story. Net revenue rose 30.4% year over year to $112.3 million, extending the growth streak to 20 straight quarters. And in that same quarter, the company posted a net loss of $12.1 million, according to The Wall Street Journal’s read of the filing. A brand growing revenue 30% and posting a quarterly loss in the same three months is not a broken business. It’s a business where the gap between top-line growth and bottom-line health has widened enough that the company’s own S-1 had to explain it, in writing, to prospective public shareholders. Few private DTC operators are forced to write that sentence down for anyone.

The filing also discloses a dividend recapitalization, new debt taken on and cash paid out to existing stockholders, that happened eight days before the S-1 was filed. Retail TouchPoints flagged the timing directly as noteworthy. Dividend recaps ahead of private equity-backed IPOs aren’t unusual on their own. The detail worth sitting with is what it signals about how Permira, which holds a majority stake and will retain what the filing itself calls substantial sway over corporate decisions post-listing, is positioning its own return before public shareholders get a vote.

Why the growth-versus-margin gap matters beyond one brand’s balance sheet

Reformation is one of the very few DTC brands operating at real scale, roughly 90% of revenue direct-to-consumer, 1.14 million active customers, 70 owned stores, that has been consistently profitable since 2018. If tariffs cut 360 basis points off gross margin at a brand with this much scale, sourcing sophistication, and operating discipline, that’s a real, measurable data point for what tariff exposure is actually costing DTC apparel businesses right now, not a projection or an industry survey average. Most operators have no equivalent number for their own business, because most operators aren’t required to disclose one.

The mechanism worth understanding precisely: revenue growth and margin health can move in opposite directions at the same time, and a brand can look stronger on the metric everyone checks first, top-line growth, while getting structurally weaker on the metric that actually determines whether the business is sustainable. Reformation’s 30.4% Q1 revenue growth is a genuinely strong number. The $12.1 million net loss sitting inside that same quarter is the number that would get missed by anyone reading only the growth headline, and it’s exactly the kind of gap this newsletter has flagged before in a different context, CTR rising while conversion falls, CTR up while ROAS down. Growth and health are not the same axis, and a filing that discloses both in the same paragraph is a rare, direct look at how wide that gap can get even inside a well-run business.

What actually held margin up, and what a smaller brand can take from it

Reformation’s S-1 credits a specific operational detail with driving measurable performance inside its retail footprint: the company’s patented Retail X store format, a showroom model with one sample garment on display where customers build their order through a touchscreen rather than browsing full floor inventory. That format drives 8.5% higher average order value than Reformation’s standard stores, according to the company’s own filing, and now accounts for roughly 75% of the brand’s retail locations as of Q1 2026.

That’s a concrete, disclosed, single-number result tied to a specific format decision. The mechanism is worth understanding rather than just copying the format wholesale. A showroom model with minimal physical inventory and a digital ordering layer removes the browsing friction of a full-stock store while still giving customers the in-person fit and fabric check that drives conversion in apparel specifically. It also means a smaller physical footprint carries a larger effective catalog than its square footage would suggest, since the touchscreen ordering system isn’t constrained by what’s hanging on the rack. For a brand with any store presence at all, the useful question is whether your own physical retail is currently forcing customers to browse against limited on-floor stock when a showroom-plus-digital-catalog model could both lower inventory carrying costs and lift AOV the way Reformation’s own disclosed number shows it did for them.

What to actually check in your own numbers after reading this filing

Pull your own gross margin trend over the same period Reformation disclosed, and separate tariff-driven cost from every other input. If you can’t isolate the specific basis-point impact of tariffs on your own margin the way Reformation’s filing does, that’s a gap worth closing before your next planning cycle, not a nice-to-have. A number you can name specifically is a number you can act on. A number buried inside a blended COGS figure isn’t.

Check whether your own quarterly revenue growth rate and your own quarterly net income are telling the same story or different ones. Reformation’s filing makes it easy to see the divergence because it’s disclosed side by side in a single document. Most private operators have to go pull this deliberately, cross-referencing a growth dashboard against a P&L that often lives with someone else entirely. If nobody on your team is checking both numbers together on a standing cadence, this is the moment to start.

If you operate any physical retail, audit whether your store format is optimized for AOV or just for square footage utilization. Reformation’s 8.5% AOV lift from a showroom-and-touchscreen format is a specific, disclosed number tied to a specific structural choice, not a general argument for smaller stores. The question worth asking is whether your own retail locations are forcing a browsing experience that limits what a customer sees and buys, in a category where a leaner, digitally-assisted format has now shown a measured lift.

If your business has outside investors, understand exactly what a dividend recap or similar pre-liquidity-event transaction would mean for your own cap table before it happens, not after. Reformation’s disclosure that new debt and a cash distribution landed eight days before the S-1 filing is a detail every operator with institutional investors should be able to explain about their own company if asked, since it’s exactly the kind of transaction that becomes public and scrutinized the moment a company files.

Your action list

• Pull your last four quarters of revenue growth rate and net income side by side in one view, the way Reformation’s own S-1 discloses both. If the two lines are moving in opposite directions and nobody on your team has flagged it, that’s the first thing to address before your next budget cycle.

• If tariffs touch your COGS at all, isolate the specific basis-point or dollar impact for your most recent full year, separate from every other input into gross margin. A number this specific is what lets you actually plan around it instead of treating tariff cost as background noise.

• If you run any physical retail, pull your AOV by location format if you operate more than one type of store, and compare against what a leaner, showroom-style format might do to that number, using Reformation’s disclosed 8.5% lift as a real, if not directly transferable, benchmark to test against.

Sources: Reformation, Form S-1 registration statement, filed with the US Securities and Exchange Commission (June 25, 2026); Retail Dive, "Reformation's IPO filing shows profitable DTC is possible" (June 26, 2026); Retail TouchPoints, "Six Things to Know About Reformation's Business as the Company Prepares to Go Public"; WWD, "Reformation Files for IPO Amid Strong Growth and Sustainability Focus" (June 26, 2026); The Industry.fashion, "Reformation files for NYSE IPO as revenue tops $500m"; Fashion United, "Reformation launches IPO roadshow, targeting up to 239 million dollars" (July 20, 2026); Quartz, "Reformation files for NYSE IPO as annual revenue tops $500M," citing The Wall Street Journal's reporting on Q1 2026 net loss.

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually happening

QVC Group filed prepackaged Chapter 11 petitions on April 16, 2026, after eight months of pre-petition negotiations with three creditor groups holding claims tied to a $2.90 billion revolving credit facility, $2.15 billion in secured notes, and $1.48 billion in unsecured holding-company notes. The court approved the restructuring plan on July 15. Debt drops from roughly $6.6 billion to $1.325 billion, vendors get paid in full or reinstated, and the reorganized company gets a new $600 million revolving credit line once the remaining closing conditions clear. Strategic Value Partners and Silver Point Capital, the two largest lenders, take control of the equity.

The collapse behind that filing is a familiar one. Between 2018 and 2024, QVC’s and HSN’s main television channels each lost close to half their US audience, 44% and 47% respectively, according to Fortune’s reporting on the restructuring. Cable and satellite subscriptions kept shrinking, linear TV viewership kept falling, and a business built on shoppers discovering products by flipping channels had fewer and fewer channels left to flip past. Four straight years of losses and a revenue base pressured on every side is the plain financial story.

Buried inside that same period is a specific, measurable exception. QVC launched on TikTok Shop in August 2024 and expanded to a full 24/7 live shopping stream on the platform in April 2025. By the time the company filed for bankruptcy a year later, it had acquired close to 1 million new US customers through that single channel, more than 100,000 in Q2 2025 alone according to Modern Retail’s reporting, and generated 35 million livestream views in less than a year. QVC Group was named TikTok Shop’s Seller of the Year in 2026. None of that stopped the bankruptcy. All of it happened anyway, inside the same twelve months.

Why this matters beyond one company’s balance sheet

The instinct is to read a bankruptcy filing as evidence that a strategy failed. That’s the wrong read here, and QVC’s own court disclosures make the more precise point directly: monthly debtor financial reports filed during Chapter 11 show TikTok Shop as one of the only parts of the business still adding customers while the rest of the company was shrinking under debt service it could no longer support. The bankruptcy wasn’t caused by the pivot to social commerce. It happened despite a pivot that was, by the specific metric of new customer acquisition, working.

That distinction matters for any operator evaluating live commerce as a channel right now. QVC’s TV audience decline and its TikTok Shop customer growth were not two separate stories running in parallel. They were the same story. A retention channel, cable subscriptions, was collapsing at a rate no amount of programming could fix. An acquisition channel, live social shopping, was proving that the underlying format, a host talking through a product in real time, driving an immediate purchase decision, still works when it’s placed where the audience actually is. The channel didn’t change. The screen did.

The scale QVC brought to that transition is worth naming honestly, because it’s not fully replicable at smaller size. More than 74,000 creators have featured QVC items through shoppable videos and livestreams. The company runs five dedicated TikTok channels, over 220 live hours per week, and had 95,000 products listed on the platform as of its 40th anniversary event in September 2025. That’s a made-for-TV production operation redirected at a phone screen, not a small team improvising a livestream. A $5 million DTC brand isn’t going to replicate 220 hours of weekly live content.

What is replicable is the specific mechanism QVC’s own leadership pointed to. Alex Wellen, the company’s president and chief growth officer, told a live shopping summit that TikTok Shop was already the fastest-growing customer base across the entire company, predicting it would overtake every other channel QVC operates in the short term. Krystyna Taheri, the company’s SVP of social commerce, put the underlying logic more simply at the TikTok Shop Awards: “TikTok Shop is us.” QVC didn’t invent live shopping. It had been running televised live shopping since 1986. What TikTok Shop offered was a distribution surface where that format could reach an audience that had already stopped watching cable, without QVC needing to convince anyone that live product demonstration works as a sales mechanic. The format was proven. Only the screen needed to change.

What to actually take from this if you’re not a $9 billion legacy retailer

Don’t read “QVC filed for bankruptcy” as evidence against live commerce as a channel. Read the two facts together: the parts of the business tied to declining, subscription-dependent attention collapsed, and the part of the business built around live, creator-driven, platform-native demonstration grew a customer file that had been shrinking for four years. If your brand has any product that benefits from being seen in use, demonstrated, or explained in real time, that’s the part of QVC’s story worth studying, not the debt restructuring.

Check whether your product category has a real live-demonstration advantage before assuming the channel applies to you. QVC’s strongest TikTok Shop categories track closely with what worked on television for four decades: beauty, home goods, fashion, kitchen items, categories where seeing a product used or worn changes the purchase decision more than a static photo does. A product that doesn’t benefit from being shown in motion won’t get the same lift from a live format, regardless of platform.

Treat creator volume as a distribution strategy, not a single-host bet. QVC’s model works with 74,000-plus creators tagging and selling products independently, alongside its own branded channels. That’s the part smaller brands can actually approximate: a defined product catalog available for any creator to tag and sell, rather than betting the whole channel on one paid partnership. The affiliate infrastructure matters more than any single big-name booking.

Measure new-versus-existing customers on the channel specifically, the way QVC’s own leadership did. The number that made TikTok Shop worth defending inside a bankruptcy filing wasn’t total sales. It was new customer count, tracked separately from the rest of the business, showing growth the company hadn’t seen anywhere else in four years. If you’re running TikTok Shop or any live commerce channel and only tracking blended revenue, you’re missing the metric that would tell you whether the channel is actually acquiring anyone new or just reselling to people who already knew your brand.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

How a 5,000-follower creator sold Pacsun 11,000 pairs of jeans in 36 hours, without a contract

In late 2023, Lyla Biggs, a Pacsun customer with about 5,000 TikTok followers and no brand deal, bought a pair of the retailer’s Casey Low Rise Baggy Jeans, went home, and filmed a simple styling video in her bedroom. CEO Brieane Olson told Inside Retail Asia in July 2026 that the video drove demand across every channel and sold 11,000 pairs almost immediately, with no paid promotion behind it.

The numbers: That single jean style went on to sell over 200,000 pairs. Pacsun has since sold more than 1 million pairs of jeans and denim through TikTok. The brand now has 2 million TikTok followers, runs an open creator platform where any customer can self-select as an advocate, and operates a 50/50 organic-to-paid marketing split after bringing all paid marketing in-house. For the first time in 18 years, Pacsun is opening more new stores than it’s closing.

Why it worked: Pacsun treated the video not as a one-off viral moment but as validation of a structural bet it had already made. Olson described the company investing early in TikTok and YouTube specifically to build what she calls co-creation infrastructure, well before any single video proved the model. Because that infrastructure, an open platform where unpaid customers can tag and post, and a team ready to amplify what performs, already existed, the brand could turn one bedroom video into a sustained sales channel instead of a viral spike that faded in a week. Olson has been explicit that the distinction matters: building with customers, not just marketing to them, is what let an uncontracted creator’s post scale into over a million units sold. Lyla Biggs now sits on Pacsun’s Youth Advisory Council, a group with direct input into brand decisions, turning the company’s single best organic result into an ongoing structural relationship rather than a moment it tried to repeat from scratch.

How to replicate it:

Build the infrastructure before you need the moment. Pacsun’s open creator platform existed before Lyla Biggs posted. A brand scrambling to set up creator tagging and amplification after something goes viral misses the window a pre-built system would have caught immediately.

Let customers self-select rather than requiring a contract to participate. The video that sold 11,000 pairs in 36 hours came from an unpaid fan, not a bookable partnership. A program that only amplifies contracted creators structurally excludes the exact kind of authentic post that converts best.

When something organic performs, formalize the relationship instead of treating it as a one-time hit. Pacsun didn’t just reshare Lyla Biggs’s video. It brought her into an ongoing advisory role, turning one result into a repeatable source of insight.

Track how much of your best-performing content originates from actual customers versus brand-produced assets. Pacsun’s benchmark, a majority of the vertical video on its own site now comes from real users, is a concrete target smaller brands can measure themselves against.

Your action list

• If you sell any product with a genuine show-me-how-it-works or see-it-on-a-body advantage, check your current live commerce presence, or lack of one, against that specific product set rather than your full catalog. Live format lift concentrates in categories where demonstration changes the decision.

• Pull new-versus-returning customer data for any live or social commerce channel you’re already running, separate from your blended revenue number. If you can’t currently isolate that number, that’s the first fix, since it’s the one metric that would tell you whether the channel is genuinely acquiring anyone new.

• If you’re not currently enabling third-party creators to tag and sell your products independently of paid partnerships, look at what it would take to open that up. QVC’s acquisition volume runs through tens of thousands of creators working the catalog on their own, not through a small number of expensive bookings.

Sources: Inside Retail Asia, "Pacsun CEO Brieane Olson discusses how the brand taps into the power of co-creation," July 2026; Modern Retail Podcast, "Pacsun's CEO on how the brand cracked its Gen Z strategy," May 2026; Bloomberg Businessweek via BigGo Finance, "PacSun 'Co-Creating' Brand to Reach Gen Z", QVC Group, "QVC Group Achieves Key Milestone with Court's Approval of Comprehensive Financial Restructuring Plan" (July 15, 2026); Retail Dive, "QVC Group nears bankruptcy exit with approved restructuring plan"; Digital Commerce 360, "QVC Group restructuring approval bankruptcy exit" (July 17, 2026); DTC Dispatch, "QVC Group Gets Court Approval to Exit Bankruptcy With $5.3 Billion Debt Cut" (July 17, 2026); Vista Today, "Judge Rejects Shareholder Challenge, Clears Path for QVC's Bankruptcy Exit"; Forbes, "QVC Was Slow To The Shift, And Now It's Costly To Catch Up" (April 21, 2026); Modern Retail, "'Social scrolling is the new channel surfing': Behind QVC's TikTok strategy"; Fortune, "TikTok is the centerpiece of QVC's comeback strategy" (March 25, 2026); Retail Dive, "QVC celebrates 40th anniversary with TikTok Shop event"; QVC Group, "QVC Group is named 'Seller of the Year' at TikTok Shop Awards"; Inside Retail Asia, "Pacsun CEO Brieane Olson discusses how the brand taps into the power of co-creation" (July 2026); Modern Retail Podcast, "Pacsun's CEO on how the brand cracked its Gen Z strategy" (May 2026).

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What happens if your product data streams through the legacy API? Simple: your listings drop off Google Shopping instantly.

That is a genuine business threat. Google Shopping accounts for 76% of retail search ad spend, with Q1 2026 data showing an 18% year-over-year growth rate (according to Smarter Ecommerce). Going dark for 48 hours in August will ruin your quarterly sales numbers.

Check if you are actually exposed before doing anything else.

Checking your account takes 30 seconds

Open Google Merchant Center. Navigate to Settings, click Data Sources, and inspect the Source column.

If that column reads “Content API”, you are at risk. You must migrate before August 18.

If it displays a flat file like CSV or XML, a scheduled fetch, a Google Sheet, or a platform app name, you can breathe easily. You are safe. The shutdown targets programmatic feeds running on the old API infrastructure.

That filters out the vast majority of standard stores. Even so, I spoke with two brand managers last week who assumed they were fine until someone looked. They found custom scripts built back in 2022 still piping catalog updates through the legacy connection.

How to respond based on your infrastructure

Standard e-commerce platforms: Shopify, WooCommerce, and BigCommerce merchants are mostly covered. Official sales channel plugins updated to Merchant API v1 earlier this year. Still, verify your Merchant Center dashboard to confirm “Content API” is absent from that Source column. Assumptions cost money.

In-house scripts and custom integrations: This is where real exposure sits. Custom code built on the old Content API needs a full rebuild on Merchant API v1. Google estimates complex migrations take 16 to 20 weeks including QA testing. That timeline is gone. You are in rapid triage now.

Get a developer assigned today. If internal dev capacity is tight, consider middleware like DataFeedWatch, Channable, or Feedonomics. These platforms connect to Merchant API v1 out of the box, saving your engineers from writing custom code under deadline pressure.

Watch out for feed label mapping. Many marketers forget this detail. If your Google Ads campaigns rely on feed labels to organize products into PMax asset groups or Standard Shopping campaigns, breaking those label linkages during migration will stop your ads from serving. Ensure feed label verification is explicitly included in your developer brief.

The opportunity hidden behind the deadline

Fixing your API connection handles compliance. What you do with your feed afterward determines whether you win market share.

The new Merchant API transmits inventory and pricing changes much faster. Speed matters when running Performance Max or Standard Shopping with automated bidding strategies. Fresh feeds feed better signals into the algorithm. Out-of-stock items shown as available waste ad spend and trigger data quality penalties over time.

Upgrading your feed pipe is step one. Most brands will stop there, mark the ticket complete, and wonder why campaign ROI stays flat.

That is a mistake. Smarter Ecommerce analyzed over 3,000 PMax campaigns in 2026 and found smart bidding algorithms have reached maturity. Bid strategy is no longer the main driver of performance. Feed quality and asset group relevance are.

If your brand spends $50,000 monthly on Google Shopping, you are likely losing more revenue to poor feed attributes than to imperfect bid targets.

A single optimized feed feeds four channels simultaneously: paid Shopping ads, PMax campaigns, AI Mode recommendations (which hit 75 million daily active users in January 2026, according to Google data cited by Digital Commerce 360), and Universal Commerce Protocol for agentic checkout. Clean up your feed once, and all four channels benefit.

Five high-impact feed improvements

1. Prioritize your top 20% revenue SKUs
Forget about optimizing your entire 10,000-product catalog at once. Pull your sales reports and isolate the top quintile. These products already possess conversion history and auction volume, meaning feed adjustments yield measurable performance shifts in days. Work on the long tail next month.

2. Rewrite product titles using Search Terms Reports
This is the highest-ROI feed edit you can make. EasyApps 2026 performance data shows structured titles generate 15% to 30% more impressions and 10% to 20% higher click-through rates compared to generic product names, without increasing your bids.

Open Google Ads, pull the Search Terms Report for the last 90 days, and sort by conversions. Identify buying queries missing from your current titles and insert them. Use a logical structure: Brand + Product Type + Key Attributes (color, size, material, model number). Keep titles under 150 characters, placing essential attributes in the first 70 characters so mobile cards do not cut them off.

3. Resolve GTIN and attribute errors
Missing Global Trade Item Numbers hurt auction performance. Products lacking valid GTINs get relegated to lower-tier auctions and ignored by AI Mode. In Merchant Center, go to Diagnostics, filter by missing attributes, export the report, and fix GTINs first. Next, populate optional fields like color, size, material, and age group. These fields are technically optional, but automated bidding systems use them as core matching signals.

4. Deploy margin-based custom labels
Bidding the same target ROAS across 60% margin items and 15% clearance stock destroys profitability. Add a custom label to your feed reflecting net margin tiers (such as high-margin, mid-margin, clearance). In Google Ads, group products by these labels. Lower your target ROAS on high-margin SKUs so Google bids aggressively where margin exists, and raise target ROAS on clearance items to protect your margins.

5. Enable AI Max on proven Standard Shopping campaigns
Google released AI Max for Shopping on April 30, 2026, as an extension for Standard Shopping. It expands campaign reach into discovery and conversational queries without forcing a total migration to Performance Max.

Only turn this on for campaigns generating at least 30 conversions per month. Below that volume, the system lacks sufficient signal. Before launching, exclude non-commercial URLs (like blog posts or login pages) and set strict brand text guidelines to prevent off-brand AI copy. Internal Google data cited in 1ClickReport’s March 2026 guide logged an average 14% conversion gain at comparable CPAs.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

Kendra Scott built 8,000 pages for AI search instead of chasing traditional keywords

Digital Commerce 360 profiled jewelry brand Kendra Scott in November 2025 after they built 8,000 site pages over 12 months specifically designed for AI search summaries rather than keyword density.

Instead of optimizing purely for short terms like “14k gold ring”, they built content around long-tail concepts, gifting intent, and specific wear cases.

The results were striking. Those pages captured 5% of Kendra Scott’s annual web traffic, with 27% ranking on page one of Google Search results.

Kamanasish Kundu, SVP of digital and e-commerce at Kendra Scott, explained to Digiday that the team prioritized concepts and use cases engineered to perform in AI summaries rather than matching basic category phrases.

This approach applies directly to your product feeds. AI Mode reads your Merchant Center product descriptions the same way. A basic title like “Gold Solitaire Ring” matches keyword lookups, but a detailed description explaining that the ring is a popular first-anniversary gift, pairs with specific bangles, and fits true-to-size answers conversational prompts.

Google has not revealed the exact weight AI Mode assigns to descriptions versus titles. However, search results clearly favor product data that reads like a helpful answer over a raw spec sheet. Take 30 minutes this week to rewrite descriptions for your top 10 products to answer conversational buyer questions.

Your action list

• Run the 30-second audit: Settings > Data Sources > Source column inside Google Merchant Center. Confirm whether “Content API” is present.

• Update titles: Pull your 90-day search term report, find the 10 highest-converting queries missing from product titles, and add them.

• Apply margin labels: Add a custom margin label to your top revenue SKUs, adjust product group targets in Google Ads, and evaluate blended ROAS over 30 days.

Sources: Smarter Ecommerce analysis of 3,000+ PMax campaigns, cited in Scubemarketing, "Google Shopping Statistics 2026" (July 2026); AI Advantage Agency, "Google Merchant API Migration" (May 2026); AI Advantage Agency, "Google Shopping for Ecommerce: Complete Strategy Guide 2026" (May 2026); AI Advantage Agency, "Product Feed Optimization for Ecommerce" (May 2026); Sitation, "The Google Merchant API Deadline" (July 2026); Digital Applied, "Merchant Center Drops Next" (July 2026); XICTRON, "Google Merchant API Migration: What Shop Owners Need to Know by August 2026" (January 2026); PPC News Feed, "Google Shopping API Migration Deadline Coming Soon" (January 2026); ALM Corp, "Google Shopping API Migration Deadline: February 28 and August 18, 2026 Guide" (January 2026); 1ClickReport, "Google Ads AI Max for Search 2026: Full Setup Guide" (March 2026); Groas.ai, "Google AI Max: The 2026 Master Guide" (May 2026); AI Advantage Agency, "Google Shopping for Ecommerce" citing AI Mode daily active users from Google data via Digital Commerce 360 (February 2026); EasyApps merchant performance data cited in Passionfruit, "Google Shopping Feed Optimization" (April 2026); Digital Commerce 360, "Ecommerce Trends: How online retailers are preparing for Google Zero" citing Kendra Scott and Kamanasish Kundu via Digiday (November 2025); AI Advantage Agency, "Google Shopping for Ecommerce," August 18 deadline analysis (May 2026).

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

At the same time, the U.S. Trade Representative’s accelerated Section 301 investigation, the administration’s replacement mechanism proposing 10 to 12.5 percent duties on 60 trading partners with no statutory expiration date, was given a completion deadline of today, July 20. Whether it lands on time, slips by days, or gets delayed by litigation determines what every import shipment costs starting Thursday morning.

If you run a brand that sources from China, Vietnam, India, South Korea, Japan, or the UK, this is the most operationally significant four-day window of 2026. Note: EU goods including Germany moved to a separate 15% all-inclusive ceiling under the EU-US trade deal effective July 1, 2026, and are already insulated from the July 24 expiry. The decisions that need to happen before July 24 include entry timing on in-transit shipments, pricing adjustments, Q3 inventory assumptions, and HTS classification audits that should have started six weeks ago. The decisions that happen after July 24 depend on which of three scenarios actually materializes.

This issue covers what is actually happening, what the three scenarios mean for your P&L, and what to do before Thursday.

What Section 122 actually was, and why this moment matters

To understand why July 24 matters, you need to understand the legal chain that produced it.

On February 20, the Supreme Court ruled 6 to 3 in Learning Resources, Inc. v. Trump that the International Emergency Economic Powers Act does not authorize the president to impose tariffs. That ruling struck down the entire IEEPA tariff architecture that had governed U.S. trade policy since early 2025, including the stacked tariffs on Chinese goods that had pushed some effective duty rates above 100%.

Four days later, on February 24, the administration imposed a 10% global import surcharge under Section 122 of the Trade Act of 1974. Section 122 is a balance-of-payments emergency authority. It allows a surcharge of up to 15% for a maximum of 150 days. No presidential extension is possible. Hard statutory cap. Day 150 is July 24.

The U.S. Court of International Trade ruled on May 7 that this surcharge was also unlawful. That ruling granted a permanent injunction for three specific importer plaintiffs only (the State of Washington, Burlap and Barrel, and Basic Fun); it was not a universal injunction and does not stop collections for other importers. The Federal Circuit stayed even that limited ruling, so CBP has continued collecting the duty through today. That stay does not change the July 24 sunset. The surcharge expires by statute at 12:01am Thursday regardless of how the Federal Circuit ultimately rules on the merits.

According to Capital Economics trade-weighted estimates, the average effective U.S. tariff rate under Section 122 has been running at roughly 11 to 13 percent, depending on methodology, compared to a post-IEEPA-strike-down baseline of approximately 8 percent if Section 122 lapses with no immediate replacement. That reflects a reversion to pre-IEEPA baseline rates: MFN duties plus the existing China-specific Section 301 lists that predate 2025.

That 3 to 5 point swing is not a rounding error. On a $2 million import program, it is a $60,000 to $100,000 cost structure shift in a single business day.

The three scenarios, and what each means for your margins

Scenario A: Section 301 finalizes on or before July 24.

The USTR’s proposed replacement, 10 to 12.5 percent duties on 60 trading partners proposed under the June 2 determination following forced-labor enforcement investigations, has a July 20 completion deadline. Of those 60 economies, 46 face the higher 12.5 percent rate (no prohibition on forced-labor imports exists at all) and 14 face 10 percent (prohibition exists but is not effectively enforced). If finalized on schedule, most brands see their effective tariff rate shift from 10 percent (Section 122) to somewhere between 10 and 12.5 percent (Section 301), depending on country of origin. The headline rate does not change much, but the legal authority does. Unlike Section 122, Section 301 has no statutory expiration and no rate ceiling. If you import from a country in the 60-nation tier, your duty exposure becomes permanent rather than temporarily authorized.

For operators: Scenario A means your current cost model largely holds, but the volatility premium you have been pricing in, the possibility that this all goes away, no longer applies. Start modeling cost structure assuming these rates are structural, not temporary.

Scenario B: Section 122 lapses with no immediate replacement.

If the Section 301 finalization slips due to litigation, comment-period challenges, or administrative delay, Section 122 expires Thursday with nothing behind it. Imports revert to MFN plus existing pre-2025 China Section 301 duties. For brands sourcing from Vietnam, India, Japan, or the UK, this is a brief but meaningful rate reduction. A Vietnamese-manufactured consumer goods shipment currently facing 10 percent (Section 122) could enter at 4 to 6 percent (MFN) for whatever period Section 301 takes to finalize.

This creates what trade lawyers are calling a gap window: entry dates are what matter for duty rates, not ship dates. Shipments already on the water can be timed to enter after 12:01am Thursday. According to tariff compliance guidance published last week by multiple customs brokers, this is a legal and commonly used tactic. The correct move is to call your customs broker today.

For operators: Scenario B is a short-term margin tailwind but not a strategy. Section 301, if finalized at 12.5%, lands higher than Section 122’s 10 percent for some categories. Brands that treat the gap window as a permanent reprieve will be unprepared for what comes after it.

Scenario C: Section 301 finalizes at 12.5% for your country of origin, after a brief gap.

The worst-case version for brands sourcing from the affected 60 nations. You get a brief dip to MFN rates, then permanent Section 301 duties at 12.5 percent, higher than the Section 122 rate you have been paying, with no sunset built in. The administration’s June 2 determination specifically targets excess manufacturing capacity and forced-labor enforcement, which means most high-volume sourcing countries, Vietnam, India, Cambodia, Bangladesh, South Korea, are in scope.

For operators: This is the scenario to plan against, not the one to hope away. If you have been treating the current tariff environment as a temporary disruption you can absorb and wait out, Scenario C is the confirmation that you cannot.

The operational problem most brands have right now

The tariff architecture has changed so many times in the last 18 months (IEEPA stacking, IEEPA struck down, Section 122, Section 301 proposed) that a significant share of operators are running landed cost models that have not been updated since February. That was the last clean handoff point. A lot of businesses locked in their 2026 cost assumptions when Section 122 went into effect, modeled a 10 percent duty rate as the stable floor, and moved on.

That assumption breaks Thursday morning under at least two of the three scenarios above.

The brands most exposed are the ones running a single blended duty rate across their entire catalog rather than a SKU-level landed cost model that separates country of origin, HTS code, duty rate, freight, and insurance into distinct line items. Brands that built that granularity into their operations early (the ones who audited HTS classifications in Q1 and built landed cost dashboards rather than spreadsheet estimates) are the ones who can update a single variable Thursday morning and have an accurate P&L within hours. Everyone else is going to be guessing for days.

According to data from Northstar Financial Advisory published earlier this year, brands that updated their landed cost models and pricing strategy within 90 days of the 2025 tariff changes preserved 85% of their pre-tariff contribution margin. Brands that waited six months or more to respond showed permanent margin compression, not because the tariffs were worse, but because they raised prices late, absorbed the cost during the delay, and then faced customer resistance to larger, more sudden increases on top of the initial absorption.

The operational failure mode is not misunderstanding the tariff law. It is having a cost model that lags the policy environment by 60 to 90 days. In a window like this week, where the rate structure can move by several points in either direction in a single day, a 60-day lag is the difference between knowing your Q3 margin on Tuesday and not knowing it until September.

There is also an HTS classification problem that has compounded throughout 2026. According to multiple customs compliance sources, misclassification is one of the most common and expensive errors in the current environment. With stacked tariff programs (base MFN, Section 232 on metals, existing China Section 301 lists, and now Section 122 or Section 301 on top) the difference between an accurate and inaccurate HTS code on a single product can be 15 to 20 percentage points of duty. Brands that have never had an HTS classification review are almost certainly overpaying in some categories and potentially underpaying in others.

Why the “just wait and see” posture does not work here

Every tariff transition since early 2025 has produced a cohort of operators who decided to wait for clarity before making adjustments. In each case, the clarity arrived at the same time for everyone. But the brands that had already built scenario models, audited their supply chains, and had pricing changes ready to push were able to move within days rather than weeks. The gap between ready and catching up is where margin gets permanently lost.

There are four decisions in front of every importing brand right now that cannot be made retrospectively.

Entry timing on in-transit goods. Duty rates are assessed at entry, not at shipment date. Goods currently on the water can be held at port and entered Thursday if the gap window scenario materializes, legally capturing the rate differential. This decision has to happen before Thursday morning and requires active coordination with your customs broker today.

Q3 pricing. If you have been holding price increases because you expected tariff relief after July 24, both Scenario A and Scenario C make that calculus wrong. Scenario B gives you a brief window but prices it away in six to twelve weeks when Section 301 finalizes. Pricing decisions made against a temporary disruption assumption need to be re-examined today against a structural rate environment assumption.

Q4 inventory orders. Every unit you order in the next thirty days carries a duty assumption. That assumption needs to reflect a scenario range, not a point estimate. Orders locked in against a 10% duty assumption that end up landing under 12.5% Section 301 are underwater before they arrive.

Sourcing diversification math. Every country in USTR’s 60-nation Section 301 tier gets re-evaluated by this determination. The Mexico/USMCA play (duty-free for qualifying goods, 4 to 8 day truck transit versus 25 to 35 day ocean transit from Asia) looks materially better in a Scenario A or C environment than it did in the IEEPA era. Brands that have not run the landed cost comparison with USMCA-qualifying sourcing in the last 90 days are leaving money on the table regardless of how Thursday resolves.

The July 24 playbook: what to do before Thursday at midnight

Step 1: Call your customs broker today, specifically about entry timing.

This is the highest-leverage call of the week for any brand with shipments currently in transit. Ask them three questions: Which of my in-transit shipments are scheduled to enter before July 24? Can we legally delay entry to after 12:01am Thursday? And what is the cost and operational complexity of doing so? The answer depends on your specific shipments, ports, and carrier arrangements, but the decision window closes Thursday morning. Many brokers are actively working this for clients right now. If yours has not raised it proactively, raise it yourself.

Step 2: Pull your HTS codes and run a classification audit on your top 20 SKUs by import volume.

Not your entire catalog. Start with the 20 SKUs that represent the largest share of your duty payments. For each one, verify the HTS code, the applicable duty rate under each of the three July 24 scenarios, and the total duty impact per unit. The goal is a scenario matrix, not a single number. If you do not have internal customs expertise, most brokers and trade compliance consultants can turn this around in 48 to 72 hours.

Step 3: Update your landed cost model at the SKU level, and set it up to update automatically.

Your landed cost model needs to separate at minimum: FOB factory cost, ocean freight, inbound tariffs by HTS code and country of origin, port charges, inland freight, and any applicable duty drawback. If tariff rates are a single blended line or a rough percentage applied to all COGS, you are managing your margin with a wrong instrument. Tools like Nventory, TrueProfit, or a well-structured Inventory Planner integration can build this granularity if you do not already have it. The goal is not accuracy today. It is being able to update one variable Thursday morning and know what happened to every SKU in your catalog.

Step 4: Scenario-model your Q3 contribution margin against all three July 24 outcomes.

You need three versions of your Q3 P&L: one where Section 301 finalizes at 12.5% on your key origin countries (Scenario A/C); one where rates briefly revert to MFN for 4 to 8 weeks then step up to 12.5% (Scenario B to C); and one where rates revert to MFN and stay there longer than expected. For each scenario, identify at what blended duty rate your current pricing becomes margin-negative on each category. That breakeven rate is your trigger for a price adjustment.

Step 5: Decide on your July 24 pricing posture before Thursday morning, not after.

Eighty-seven percent of DTC merchants raised prices in 2025 to 2026 to offset tariff costs, according to Yotpo’s 2026 DTC Brand Comparison. The brands that preserved margin did so by raising prices within 2 to 4 weeks of the tariff change, not by absorbing costs and raising prices months later under duress. If Section 301 at 12.5% materializes, you want your price changes ready to stage: updated on your DTC site within days, staggered on marketplace listings over 2 to 3 weeks to avoid triggering algorithmic suppression from sudden price jumps. That preparation happens before Thursday, not after.

Step 6: For any category where your tariff exposure exceeds $200K annually, run the Mexico/USMCA landed cost comparison this week.

Mexico’s nearshoring advantage is not hypothetical. USMCA-qualifying goods enter the U.S. at 0% duty. Road freight from northern Mexico manufacturing zones (Monterrey, Juarez, Tijuana) to U.S. distribution centers takes 4 to 8 days versus 25 to 35 days by ocean from Asia. Foreign direct investment into Mexican manufacturing hit $40.87 billion in 2025, meaning capacity is expanding. For brands whose China, Vietnam, or Indian origin goods are facing stacked Section 301 duties at 12.5% or higher, the USMCA arbitrage (even accounting for higher per-unit manufacturing costs) often pays back within 12 to 18 months on categories with sufficient volume. Get the quote. Run the math. The decision does not have to happen this week, but the information gathering does.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

How e.l.f. Beauty posted 25% net sales growth through a tariff environment that compressed every competitor’s margin.

When IEEPA tariffs on China-origin goods pushed effective duty rates above 100% on some categories in 2025, beauty brands that manufactured in China faced the same landed cost shock as everyone else. Most responded by either absorbing the cost hit, implementing broad price increases that damaged conversion, or cutting marketing spend to offset the COGS increase.

e.l.f. Beauty did something structurally different, and the mechanism matters more than the outcome number.

According to Yotpo’s 2026 DTC Brand Comparison and Eightx’s margin benchmarking data, e.l.f.’s advantage was not that they avoided the tariff. Their products are China-manufactured. The advantage was in what they were selling and at what price point. A $10 lipstick can absorb a 10 to 15 percent tariff-driven price increase (moving from $10 to $11.50) without materially changing the consumer’s purchase decision relative to the $45 alternative from a luxury brand. The price sensitivity curve at the $8 to $15 retail band is flatter than the price sensitivity curve at the $30 to $50 band, because the consumer’s reference frame is different. They are not comparing e.l.f. to Fenty. They are comparing e.l.f. to drugstore alternatives that faced the same cost increase.

The result, per their fiscal year 2026 earnings (ended March 31, 2026) and data referenced across multiple analyst reports: 25% net sales growth through the full IEEPA tariff period despite the same input cost wave that caused margin compression across the beauty category. The pricing power held not because e.l.f. avoided the tariff, but because their price point allowed them to pass it through at a percentage increase that felt smaller in absolute dollar terms to the consumer.

How to replicate it:

The underlying mechanic (passing through tariff costs as a percentage increase rather than absorbing them in margin) works across categories, but the tolerance for that percentage increase varies dramatically by price point and by how much white space exists between your product and the next viable alternative. Before your next pricing adjustment, map where your product sits on that consumer reference frame axis. If your price point is $15 and the nearest alternative (including trading down) is $8, your ceiling is lower than you think. If your price point is $95 and the premium alternative is $250, your ceiling is higher.

The practical exercise: run a conversion rate analysis on your last three price increases. What was the percentage change? What was the conversion impact over 30 days? Most brands have this data somewhere and have never systematically pulled it. That analysis gives you the actual price elasticity number for your specific customer base, which is the only number that matters when you are deciding whether to absorb a 250bps margin hit or pass it through to customers who may or may not notice.

Your action list

• Before midnight Thursday: Confirm with your customs broker which in-transit shipments can legally be timed to enter after 12:01am July 24. The window to act on this is today and tomorrow. Pull your top 20 SKUs by import duty paid year-to-date and verify the HTS code and origin-specific duty rate under each of the three July 24 scenarios. If you find a misclassification, the correction is worth running immediately. Overpaid duties can be recovered through a prior disclosure process.

  Have your July 24 pricing decision staged and ready, even if you do not execute it until Thursday. Decide in advance at what effective duty rate each product category requires a price adjustment, so the decision on Thursday is execution, not deliberation.

• Before August 1: Build or update your SKU-level landed cost model with three scenario columns: Section 301 at 12.5%, MFN gap window, and MFN sustained. Reconcile this against your Q3 inventory orders and identify any units already on order that are underwater in the Scenario C case. For any origin country in USTR’s 60-nation Section 301 tier, obtain at least one competitive quote from a USMCA-qualifying Mexico supplier on your highest-volume SKU. Even if you do not switch, the quote gives you a negotiating baseline with your current supplier and a genuine alternative if rates continue escalating.

Sources: Industrial Sage, "Section 122 Tariff Expires July 24, 2026: What Happens Next" (July 15, 2026); TariffsTool.com, "Section 122 Expires July 24, 2026: Rates After + 3 Scenarios" (July 16, 2026); Nakachi Eckhardt and Jacobson, "Section 122 Global Surcharge Set to Expire July 24 by Operation of Law" (July 2026); Trade Duty Refund, "July 24 Deadline: 10 to 12.5% Impact on EU and UK E-Commerce Shipments to US" (July 2026); ShipperHQ, "Ecommerce Tariffs in 2026: Current Rules and Upcoming Updates" (June 2026); Northstar Financial Advisory, "2026 Tariff Guide for E-Commerce: De Minimis, Duties, and Pricing" (2026); Yotpo, "2026 DTC Brand Comparison: The Resilience Playbook" (December 2025); Eightx, "Average DTC Gross Margin 2026: 57% Median" (June 2026); Nventory, "2026 US Tariff Changes for Ecommerce Sellers" (February 2026); 3PL Center, "Nearshoring to Mexico is Accelerating in 2026" (April 2026); Viabox, "The 10% US Tariff Surcharge Is Set to Expire July 24" (June 2026); Anderson Perrino customs advisory, "Section 122 Expiration Approaching" (July 2026); ATTN Agency, "Tariff Impact on DTC Brands: Marketing Strategy Pivots for April 2026" (April 2026); ACG Strategic Insights, "The Tariff Clock Is Running Out" (July 2026); Global Trade Alert, "Section 122 in effect: what the US tariff regime looks like now" (2026); Covington and Burling, "USTR Announces Findings and Calls for Comments in Section 301 Forced Labor Investigation" (June 2026); Gibson Dunn, "USTR Proposes New Section 301 Forced Labor Tariffs Covering Most Major U.S. Trading Partners" (June 2026); e.l.f. Beauty, Fourth Quarter and Full Fiscal 2026 Results (May 2026); Cherry Bekaert, "Court of International Trade Finds Section 122 Tariffs Unauthorized" (May 2026); TariffsTool.com, "EU Tariff Rate 2026: 15% Deal Live as of July 1" (July 2026).

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually happening

Amazon Now runs on a network of smaller, specialized fulfillment locations placed inside dense population centers, different from the large regional warehouses powering standard Prime shipping. The service covers thousands of items, fresh produce, dairy, household essentials, personal care, electronics, delivered in 30 minutes or less, in most areas around the clock. Pricing is structured to nudge behavior: Prime members pay a flat $3.99 per order, non-Prime shoppers pay $13.99, with additional small-order fees under $15, a gap wide enough to function as a real Prime-conversion lever, not just a delivery fee.

This isn’t Amazon’s first move into fast fulfillment, and understanding the sequence matters. The company has been layering speed tiers for two years: same-day and next-day delivery expanded into more than 10,000 cities and towns, 1-hour and 3-hour delivery now covers over 90,000 items in hundreds of cities, and Prime Air drone delivery runs sub-60-minute drops in nine US locations. Amazon Now is the fastest and narrowest tier yet, essentials and perishables only, but it’s also the one built specifically around the shopping occasion competitors have been slowest to contest: the moment a customer realizes they’re out of something right now, not planning a purchase in advance.

That’s the specific mechanism behind Jassy’s basket-size claim, and it’s worth sitting with rather than skimming past. A same-day perishables order isn’t a smaller, faster version of a normal Amazon order. It’s a structurally different shopping occasion, closer to a convenience-store run or a fill-in grocery trip than a planned online purchase, and those occasions have always carried higher basket sizes and lower price sensitivity than planned shopping, because the customer is optimizing for immediacy, not for finding the best deal. Amazon didn’t build Amazon Now to compete on delivery speed for its own sake. It built it to capture a category of purchase occasion, unplanned, urgent, replenishment-driven, that Amazon’s traditional one-to-two-day model was structurally unable to touch.

Why this matters beyond “Amazon got faster”

If your brand sells in a category Amazon Now already covers, groceries, household essentials, personal care, over-the-counter health items, batteries, pet supplies, the competitive question isn’t whether Amazon Now affects your Amazon storefront sales specifically. It’s whether it’s reshaping the purchase occasion your category depends on. A customer who runs out of laundry detergent, batteries, or a specific vitamin at 8pm on a Tuesday used to have three real options: drive to a store, order from Amazon and wait a day or two, or go without until the next planned shopping trip. Amazon Now collapses that decision into a fourth option that beats all three on convenience for a huge share of those moments.

That has a direct implication for any DTC or omnichannel brand whose growth strategy has leaned on subscription and replenishment mechanics to smooth out exactly this kind of purchase occasion. If Amazon Now becomes the default answer to “I’m out of X right now” for a meaningful share of your addressable customers in Amazon Now markets, some portion of what used to convert into a subscription renewal or a planned reorder on your own site instead converts into an impulse 30-minute Amazon purchase, at a moment your brand’s own retention mechanics never had a chance to intervene.

There’s a second, quieter implication tied to Amazon’s own basket-size data. If same-day perishables shoppers are building baskets nearly three times larger and spending 80% more, Amazon has every incentive to expand Amazon Now’s category coverage over time, starting with perishables and household essentials, moving toward anything that plausibly fits an “I need this right now” occasion. Electronics accessories, health and wellness basics, beauty essentials, and pet care are all categories analysts and reporting on the rollout have flagged as natural next steps, not because Amazon has confirmed specific expansion plans, but because the underlying basket economics reward it. A brand whose product sits comfortably in a replenishment or emergency-purchase pattern today should treat “will Amazon Now cover my category eventually” as a real planning question, not a hypothetical.

What to actually do if your category is exposed

Map your product against the “right now” occasion honestly. Not every consumable product is an Amazon Now target. A specialty supplement with a considered purchase cycle behaves differently than a common household item a customer runs out of unexpectedly. Be honest about where your specific SKUs sit on that spectrum, since the brands most exposed are the ones selling genuine emergency-replenishment items in categories Amazon Now already touches or plausibly will.

Treat your subscription and reorder flow as a race against the 30-minute alternative, not just against forgetting. Traditional subscription retention logic assumes the main competitor to a renewal is the customer simply not reordering. Amazon Now introduces a faster, lower-friction alternative sitting right at the moment a subscription would normally kick in. If your reorder reminder lands the day after a customer already impulse-bought a replacement on Amazon, the mechanics of your retention flow, not just its existence, need reexamining.

Reconsider what “convenience” means in your own value proposition. For years, DTC convenience meant not having to go to a store. Amazon Now resets that bar to not having to wait even a day. Brands that have leaned on convenience as a differentiator without genuinely fast fulfillment of their own need a clearer answer to why a customer should choose a one-to-three day delivery experience over a 30-minute one, whether that answer is formulation quality, price, brand loyalty, or something else specific and defensible.

If you’re not currently selling through Amazon at all in an exposed category, revisit that decision with this data point specifically in mind. The basket-size and spend uplift Jassy described is a real, stated reason Amazon Now exists, and it’s a reasonable proxy for how much upside sits in that purchase occasion generally, not just on Amazon’s own platform. If a meaningful share of your category’s demand is shifting toward instant-gratification fulfillment, the question isn’t only whether to compete with Amazon Now, it’s whether your own fulfillment and retention strategy is built for the purchase moment customers increasingly expect, regardless of which platform ends up capturing it.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

How PlayMonster spotted a demand spike in a single Google Trends chart and turned it into a new product line in eight weeks

PlayMonster, the toy manufacturer behind Koosh balls, partnered with Wham-O in February 2026 to relaunch Hacky Sack, expecting a modest nostalgia bump. CEO Jonathan Berkowitz told Modern Retail he expected “a lot of success.” What he didn’t expect was what happened next: sometime in late May, he logged onto Google Trends and watched search volume for “hacky sack” start climbing in a way that looked less like a trend and more like a spike.

The numbers: From April 15 to July 15, US Google searches for “hacky sack” rose 1,700% year over year. On TikTok, the hashtag #hackysacks climbed 32,400%, and platform-wide searches for the term rose nearly 7,000% since the start of the year. Hacky Sack content generated more than 100 million video views across TikTok and Instagram. PlayMonster’s own product drops sold out on TikTok in under three hours, with Amazon pre-orders moving almost as fast. The company sold out of its branded inventory online by the end of May and stayed at low stock for roughly eight weeks while air-freighting in replacement product as fast as it could.

Why it worked: Berkowitz’s team didn’t wait for the trend to show up in sales data before acting. The Google Trends spike itself was the trigger. The week he noticed it, PlayMonster started developing a new Hacky Sack product line, including a light-up version and one that audibly counts kicks, built specifically to extend a moment the team correctly read as a genuine platform-driven cultural spike rather than a minor seasonal bump. Berkowitz described the read as deliberate from the start: the team saw Hacky Sack as visual, skill-based, community-driven, and shareable, exactly the shape of content that spreads on TikTok, and leaned into that from the February relaunch rather than treating social media as an afterthought bolted onto a traditional toy launch.

How to replicate it:

1. Treat a leading indicator, like a sudden search-volume spike, as a legitimate trigger for action, not just a curiosity to note and revisit later. By the time sell-through data confirms a trend, the fastest-moving competitors are already six to eight weeks ahead.

2. Build your supply chain with a fast-response lever available, even if it’s expensive. Air-freighting replacement inventory instead of waiting for standard ocean freight cost PlayMonster more per unit, but it kept product available during the exact window the trend was peaking.

3. When a product catches unexpected momentum, don’t just restock the original. PlayMonster used the surge as a signal to fast-track genuine product extensions, more than one new SKU, which extends a spike into a longer sales cycle instead of just meeting the original demand and watching it fade.

4. Match your channel strategy to how the demand is actually spreading. A product going viral on TikTok specifically needs a TikTok-native response, active presence, direct fan engagement, drops timed to the platform, not a generic multichannel campaign applied evenly everywhere.

Your action list

• Pull your product catalog and flag every SKU that plausibly fits an “I need this right now” replenishment occasion, groceries, personal care, household basics, health essentials, then check whether Amazon Now already operates in your customers’ top five geographic markets.

• Review your subscription and reorder reminder timing against how long a customer would realistically wait before an unplanned repurchase. If there’s a multi-day gap between when a customer might run out and when your system prompts a reorder, that gap is exactly where a 30-minute alternative wins the purchase.

• If your category sits close to Amazon Now’s current or likely future coverage, treat this as a strategic planning input for the next two quarters, not just a competitive news item. Whether the answer is faster fulfillment, deeper loyalty mechanics, or a clearer non-speed value proposition, the moment to decide is before Amazon Now reaches tens of millions of customers by year-end, not after.

Sources: Modern Retail, "How toy manufacturer PlayMonster helped engineer a Hacky Sack revival," July 2026, Amazon, "Amazon Now: 30-minute delivery in dozens of cities" (aboutamazon.com, updated June 4, 2026); CNBC, "Amazon launches ultrafast 30-minute delivery in dozens of US cities" (May 12, 2026) and "Amazon takes further aim at the grocery market with expanded 30-minute delivery" (May 15, 2026); TechCrunch, "Amazon launches 30-minute delivery across the US" (May 12, 2026); Supply Chain Dive, "Amazon expanding 30-minute delivery service to more cities"; Newsweek, "Everything Amazon Is Doing to Get Your Orders Faster in 2026" (May 13, 2026); Modern Retail, "How toy manufacturer PlayMonster helped engineer a Hacky Sack revival" (July 2026).

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually in the draft

The core structural change is scope. The 2019 law regulated platforms and the merchants selling on them. The new draft, a 20-article amendment according to SAMR’s own published statement, pulls in a wider set of participants across the platform economy, not just the marketplace operator and the seller, extending regulatory reach into gray areas the original law didn’t anticipate: livestreaming commerce, short-video selling, and other business models that didn’t meaningfully exist when the 2019 law was written.

Platform accountability tightens alongside the wider scope. The draft adds regulatory tools that sit alongside existing penalties like fixed fines and business suspension, rather than replacing them, according to Bloomberg’s reporting on the proposal. This builds directly on two things that already happened. China’s Anti-Unfair Competition Law was revised in June 2025 and took effect in October, and for the first time explicitly defined platform operators as gatekeepers responsible for fair conditions across their marketplace. Then, starting in February 2026, new rules barred dominant platforms from forcing sellers into promotions they didn’t choose, targeting the subsidy wars that have squeezed merchant margins across Meituan, JD.com, and Alibaba’s Ele.me throughout 2025 and 2026. Industry analyst Liu Dingding told Global Times the E-Commerce Law amendment is the next layer in a deliberate sequence: unfair competition law handled seller conduct, the February rules handled forced discounting, and this amendment now rewrites the foundational statute defining who the actors in a platform transaction actually are and what they owe each other.

The cross-border provisions are the part that reaches furthest outside China. The draft explicitly includes language supporting what regulators describe as orderly overseas expansion for Chinese platforms and protecting their legal rights both domestically and abroad. Multiple outlets reporting on the draft, including Global Times, read this as Beijing signaling it intends to actively contest regulatory pressure on Chinese cross-border platforms in the US and EU rather than quietly absorb it. Zhu Keli of the China Institute of New Economy told Global Times the amendment explicitly promotes aligning Chinese e-commerce rules, management practices, and standards with international counterparts, framed as reducing the compliance burden Chinese firms face when operating under different rules in every market they enter.

Why this matters even though it’s still just a draft

The instinct is to file this under “foreign regulatory news” and wait for a final version months from now. That instinct misses two things.

First, comment periods in China routinely shape the enforcement texture of a law even when the core text doesn’t move much between draft and final. Where the pressure lands during a 30-day window, from platforms, from industry bodies, from foreign business associations, is often visible in what changes and what doesn’t, according to the legal and policy commentary tracking this draft. If your business has meaningful exposure to Chinese platforms, whether as a buyer, a seller, or a competitor, this is the window where that exposure is easiest to influence or at least anticipate, not after passage.

Second, and more immediately relevant if you’re not planning to file a comment: this draft is happening in the same year China’s e-commerce sector crossed 15.7 trillion yuan, roughly $2.16 trillion, in transaction volume, more than 30% of the country’s total retail sales. That scale means whatever rules eventually take effect here don’t stay contained to the Chinese domestic market. Alibaba, JD.com, Pinduoduo, Meituan, and Douyin route enormous transaction volume through the exact platform structures this law defines, and the cross-border arms of that same ecosystem, Temu, Shein, AliExpress, are the supply and competitive pressure most Western e-commerce operators already deal with daily, whether they’re sourcing through them or competing against their pricing.

The sequencing also tells you something practical about what’s coming even before final text exists. Regulators moved from conduct rules (unfair competition law) to specific practice bans (coerced promotions) to now rewriting the foundational statute itself. That’s not how a government drafts something as a symbolic gesture. It’s how a government builds enforcement infrastructure in layers, each one giving the next one more legal footing. If the pattern holds, the practical bite of this amendment shows up less in the headline text and more in the implementing regulations and enforcement actions that follow it into 2027, the same way the Anti-Unfair Competition Law’s gatekeeper language became concrete only once the February 2026 anti-coercion rules gave it teeth.

What this means depending on your exposure

If you source products through 1688, Alibaba, or Chinese manufacturers directly: the amendment’s platform-accountability provisions could shift how disputes, quality claims, and supplier conduct get handled at the platform level, potentially changing your recourse when something goes wrong with a supplier relationship. Worth watching the comment period for any language specifically addressing B2B sourcing platforms versus consumer marketplaces, since the draft’s scope language suggests broader coverage than the 2019 law’s platform-and-merchant frame.

If you compete against Temu, Shein, or AliExpress in your own market: the “orderly overseas expansion” and rights-protection language is a signal, not yet a mechanism, that Beijing intends to back these platforms more assertively against foreign regulatory pressure, tariff actions, and market-access restrictions. That doesn’t change your competitive position today, but it’s a data point worth factoring into how durable you expect current tariff and de minimis pressure on these platforms to be over the next 12 to 18 months.

If you run any part of your supply chain through Chinese platforms and want a voice in this: SAMR’s public consultation is open through August 4. Industry bodies and foreign business associations typically file positions during exactly this kind of window, and if you’re part of a trade association with China-facing supply chain interests, this is the moment that association’s position gets shaped, not after the law passes.

If none of the above applies directly: treat this as an early signal rather than an action item. The regulatory sequencing pattern here, conduct rules, then specific bans, then foundational rewrite, is worth remembering as a template, because it’s a similar sequencing shape to how US and EU platform regulation has moved over the past two years. Watching how it plays out in China’s fastest-moving, largest e-commerce market is a reasonable way to get a preview of arguments and mechanisms that tend to show up elsewhere within a year or two.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

How AG1 turned “no” into a $600 million business, then knew exactly when to start saying yes

AG1, the greens powder brand formerly known as Athletic Greens, spent 14 years selling exactly one product through exactly one channel: direct-to-consumer only, no retail, no product line extensions. CEO Kat Cole, who took over from founder Chris Ashenden in July 2024, told Retail Brew at NRF in January 2025 that the discipline behind those years of no wasn’t caution. It was the strategy.

The numbers: AG1 grew from $160 million in annual revenue in 2021, when Cole joined as president and COO, to a projected $600 million in 2024, the year it turned profitable, all on a single SKU sold through a single channel. The company raised $115 million in January 2022 at a valuation over $1.2 billion. Only after hitting that scale did AG1 begin a deliberate, sequenced expansion: Costco nationwide in May 2025, Amazon and The Vitamin Shoppe shortly after, a sleep supplement called AGZ, and by April 2026 a launch into every Target store and target.com, doubling AG1’s retail door count in a single move.

Why it worked: Cole was explicit that the years of restraint bought the company something specific: a well-understood, loyal customer base and a set of internal capabilities that weren’t ready for complexity earlier. Before expanding, AG1 spent years on invisible groundwork, moving production out of New Zealand to build US supply chain capacity, running clinical trials to back product claims with real research, and hiring retail-specific talent, including a CMO who’d previously scaled Yeti. “The power in our decisions for years was saying no to those things,” Cole told Retail Brew, “but after years of behind-the-scenes work to prepare for this moment... it’s ready to start saying yes.” When AG1 did expand into mass retail, it hit a problem DTC brands often underestimate: a website can carry pages of clinical research, but a Target shelf gives you a few square inches, forcing the brand to relearn how to communicate its value in three ingredient-level claims instead of a novella of science.

How to replicate it:

1. Treat single-channel or single-SKU focus as a deliberate capital-building phase, not a limitation to escape as soon as possible. AG1 used its DTC-only years to build subscriber loyalty and understand its customer deeply before spending a dollar on retail complexity.

2. Before any retail or channel expansion, audit whether your operational backbone, supply chain, certifications, packaging, claims substantiation, can actually support a shelf environment, not just a checkout page. AG1 spent real time and money on this before its first Costco pallet ever shipped.

3. When you do move into mass retail, rebuild your messaging for the space you’ll actually have, not the space you’re used to. A DTC product page and a shelf-facing package are different communication problems, and AG1 rewrote its claims hierarchy specifically for Target rather than shrinking its website copy.

4. Sequence expansion deliberately. AG1 didn’t launch retail, new products, and new markets simultaneously. It moved channel by channel, Costco, then Amazon, then Vitamin Shoppe, then Target, giving each launch room to be evaluated on its own before adding the next variable.

Your action list

• If your business sources through Chinese B2B platforms or competes directly against Temu, Shein, or AliExpress, assign someone to track the SAMR comment period through August 4 and flag any provisions specifically addressing your exposure, sourcing platform accountability, cross-border seller obligations, or market-access countermeasures.

• Don’t treat the more expansive reporting on this draft, longer article counts, gig worker provisions, specific trade-countermeasure articles, as confirmed until it’s corroborated by SAMR’s own materials or a wider set of independent outlets. The reliable baseline right now is a 20-article draft, released July 4, comment through August 4.

• If you’re part of a trade association or industry body with China-facing supply chain interests, check whether that group plans to file a position during the comment window. That’s the easiest, lowest-effort way for an individual operator to have any influence on a law this size.

Sources: Modern Retail, "AG1 lands in Target as its year of retail expansion continues," April 2026; Retail Brew, "After 14 years selling one product in DTC, AG1 is ready to expand," January 2025; Forbes, "AG1 Scoops Up $600 Million In Revenue In 2024," December 2024, Bloomberg, "China Proposes Expanding E-Commerce Law to Cover More Digital Businesses" (July 4, 2026); Global Times, "China solicits public opinion on amendments to e-commerce law, with newly added countermeasure provisions emerging as key focus"; Xinhua, "China seeks public feedback on draft amendment to e-commerce law" (July 4, 2026); CGTN, "China releases draft amendment to e-commerce law to strengthen platform economy regulation"; China Gateway 360, "China's E-Commerce Law Overhaul: Gig Worker Protections, Trade Countermeasures, and What Foreign Platforms Must Know" (secondary source, figures not independently corroborated); Modern Retail, "AG1 lands in Target as its year of retail expansion continues" (April 2026); Retail Brew, "After 14 years selling one product in DTC, AG1 is ready to expand" (January 2025); Forbes, "AG1 Scoops Up $600 Million In Revenue In 2024" (December 2024).

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now

What’s actually happening

Adobe Analytics tracks aggregate online spend across all US retailers by counting site visits and transactions, roughly a trillion retail site visits and 100 million SKUs feed its model. By that measure, Prime Day 2026 broke every record: $8.3 billion on day one alone, more than Americans spent on Thanksgiving Day 2025, and a reaffirmed $26.3 to $26.4 billion four-day total once final numbers came in. Numerator measures something different: actual order-level and household-level data from a panel of verified shoppers. By that measure, the average Prime Day order was $45.94 to $48.36 depending on which day of the event you check, down from $54.78 to $58.37 at the same point in 2025. Average household spend across the event fell to roughly $121, down from $140 the year before.

Both are correct because they’re not contradicting each other. More total dollars came in, and each individual order carried less value. The reconciliation sits in order frequency: 58 to 59% of shopping households placed two or more separate orders during the event, essentially flat with 2025’s 58%, so total order volume grew even as the value of each order shrank. Amazon pulled in more total spend by getting more transactions out of the same shoppers, not by getting each shopper to spend more per visit.

The product mix explains why. Numerator’s day-one top sellers were Premier Protein shakes, Hefty Ultra Strong trash bags, and Liquid I.V. hydration packets. Nearly two-thirds of items purchased sold for under $20. Adobe’s category data tells the same story from a different angle: personal hygiene products rose 130% versus average June daily sales, household goods like trash bags and detergent rose 65%, baby products across the board, formula up 75 to 90%, diapers and wipes up 75 to 85%, strollers up 195 to 220%. This wasn’t a shopper racing to buy a television. It was a shopper using a four-day discount window to restock the pantry and the diaper bag. Big-ticket categories weren’t absent, Adobe logged electronics up around 105% on the deepest discounts of the event and appliances up roughly 95%, but the overall basket composition tilted toward replenishment over discretionary splurging, and that tilt is what pulled the per-order average down even as total volume climbed.

Buy Now, Pay Later usage adds a second data point pointing the same direction. BNPL orders grew 9.5% year over year to $2 billion, or roughly 6.6% of all Prime Day online orders. Shoppers weren’t just buying smaller baskets, a meaningful share of them were also financing those smaller purchases rather than paying outright, which is a value-seeking behavior showing up alongside the trade-down in basket composition, not instead of it.

Why this matters more than the headline number

If you planned Q3 inventory, ad budgets, or hiring off a “Prime Day was up 9%” read, you planned off the wrong half of the data. A 9% increase in total spend built on more orders at meaningfully lower value has different operational consequences than the same 9% built on stable order counts at higher value per order, even though the top-line growth rate looks identical on a slide.

The mechanism worth internalizing: variable order costs, pick, pack, ship, and return reserve, don’t scale down linearly with order value. A pick-and-pack fee is close to the same whether the order is worth $58 or $46. That means an 11 to 17% drop in AOV translates into a sharper drop in per-order contribution margin than the percentage suggests, according to Prime Day margin analysis from Nova Analytics. If your PPC bids, ad allocation, or promotional depth were calibrated against last year’s AOV, and most sellers set these plans months in advance based on prior-year benchmarks, a meaningful share of Prime Day 2026 orders likely cleared at breakeven or below on a fully loaded basis, even in a year where total revenue looked healthy on the dashboard.

There’s also a genuine demand-quality signal here, separate from the margin math. Consumers didn’t stop shopping. They shopped more often, compared prices more aggressively (57% of surveyed Prime Day buyers checked prices at other retailers before purchasing, per Numerator’s survey data), and spent that comparison-shopping energy on staples and replenishment rather than discretionary upgrades. That’s a specific, readable pattern about where consumer confidence sits heading into the back half of the year, and it’s a materially different planning input than “consumer spending grew 9%” taken at face value.

What to actually check in your own numbers

Recompute your Prime Day contribution margin at actual AOV, not planned AOV. If your promotional and ad budgets were sized against 2025’s average order value, or an assumed flat year-over-year AOV, pull your actual per-order economics from the event and compare. The gap between planned and actual AOV is the number that tells you whether your Prime Day allocation was profitable or merely high-volume.

Check whether your bundle and multi-pack SKUs held their attach rate. In a trade-down environment, a 2-pack that sells at a 1-pack’s effective unit economics because the bundle didn’t get its normal incremental lift is a quiet margin leak. Cross-reference bundle sell-through against your standalone SKU sell-through for the same event window.

Separate your category performance by discretionary versus replenishment before drawing conclusions. If your catalog sits in a discretionary category, apparel outside of basics, home decor, electronics accessories, this Prime Day’s overall softness in per-order value likely hit you harder than a brand selling household staples or personal care basics, regardless of how your specific promotion performed. Don’t benchmark your results against the aggregate 9% headline; benchmark against your category’s actual trade-down exposure.

Treat the return wave as part of the read, not an afterthought. Post-event returns typically land two to four weeks out. A Prime Day that looked acceptable on July 1 can look different by the time returns settle in late July, and that lag is exactly when H2 inventory and ad decisions are usually already locked in. Wait for the return data before finalizing conclusions you’re building into Q3 plans.

Use this as a genuine early read on H2 consumer behavior, not just a Prime Day post-mortem. The trade-down pattern, replenishment over discretionary spending, heavier price comparison, rising BNPL usage, is a demand-environment signal independent of Amazon’s specific event mechanics. If it holds into Q3 and Q4 planning windows, promotional depth and messaging built around discretionary upgrades will likely underperform relative to messaging built around value, restocking, and multi-unit deals.

What’s actually working

What’s actually working is a recurring section of this newsletter. Each issue, we pull one brand case study from credible journalism and reported financial data. No agency testimonials, no self-reported results. The goal is simple: one thing a real brand did, the numbers behind it, and how you can steal it at your scale.

How Amer Sports grew 32% in Q1 while barely noticing the tariff environment that’s dominating everyone else’s earnings calls

Amer Sports, the parent company behind Arc’teryx, Salomon, and Wilson, posted Q1 2026 revenue of $1.9 to $1.95 billion, up 32% year over year and well ahead of Wall Street’s $1.83 billion estimate. CFO Andrew Page told Retail Dive in a May 21, 2026 interview that the IEEPA tariff refunds working their way through the industry this year will have essentially no effect on the company’s guidance, in either direction, because tariffs never meaningfully dented the business to begin with.

The numbers: Technical apparel, anchored by Arc’teryx, grew 33% to $885 million, with direct-to-consumer sales within that segment up 41%. Outdoor performance, led by Salomon, grew 42% to $714 million, with DTC up 57%. Even Wilson’s ball and racquet sports division, the most wholesale-dependent part of the portfolio, grew 13% to $347 million. Company-wide, DTC now represents roughly half of group revenue, growing 45% overall. Full-year guidance sits at 20 to 22% growth.

Why it worked: Page was direct about the mechanism on the earnings call and in follow-up reporting: geographic and channel diversification did the work. Asia Pacific revenue grew 53%, Greater China 45%, EMEA 27%, and the Americas, the segment most exposed to US tariff policy, grew a comparatively modest 18%. No single market or channel carried enough weight to make a US-specific tariff shock a company-level problem. Arc’teryx runs over 70% of its sales through DTC, giving it direct control over pricing and margin capture that a heavily wholesale-dependent competitor doesn’t have. Wilson runs the inverse, majority wholesale, but operates in a category, racquet and ball sports equipment, with less direct China-manufacturing exposure than technical apparel or footwear. The portfolio structure meant that whatever tariff pressure did land, it never hit every brand and channel simultaneously.

How to replicate it:

1. Map your revenue by channel (DTC versus wholesale) and by source geography, not just by product category. A tariff or trade shock rarely hits every channel and market at the same intensity, and knowing where your concentration sits tells you where your actual exposure lives before a policy change forces you to find out.

2. If you’re heavily DTC in a category with pricing power, use that control deliberately. Arc’teryx’s 70%+ DTC mix let it absorb or pass through cost changes without a wholesale partner’s margin requirements sitting in the way.

3. Don’t treat tariff refunds or relief as a planning input until they actually land. Page’s approach, recognizing refunds only once received rather than forecasting their timing, is a conservative, defensible way to avoid building guidance on a legal or policy process outside your control.

4. Diversification isn’t just a China-versus-not-China question. Amer Sports’ resilience came from spreading exposure across regions, channels, and brand-level manufacturing footprints simultaneously, not from a single hedge.

Your action list

• Pull your actual Prime Day AOV and compare it to what your ad budgets and promotional depth were calibrated against. If there’s a double-digit gap, recompute contribution margin at the real number before locking any H2 spend decisions off the headline growth rate.

• Check bundle and multi-pack SKU performance against standalone SKUs for the event window specifically, since attach-rate erosion in a trade-down environment is easy to miss until margin data comes in weeks later.

• Segment your own results by discretionary versus replenishment category exposure before comparing yourself to the 9.3% aggregate figure. Your relevant benchmark is your category’s trade-down pattern, not the blended market number.

Sources: Retail Dive, “Amer Sports CFO: No visibility on tariff refunds,” May 21, 2026, reporting by Cara Salpini; corroborated by Amer Sports Q1 2026 earnings call figures), Adobe Analytics, Prime Day 2026 US online spending data, released via Retail Dive and Digital Commerce 360 (June 27 and June 23, 2026); Numerator, Prime Day 2026 Insights and Real-Time Tracker (June 23 to 26, 2026); Retail Dive, “Amazon’s Prime Day drives online sales in the US up 9.3%”; Chain Store Age, “Amazon Prime Day kicks off with declining order size, spend”; Tinuiti, “Prime Day 2026: Record Sales Meet Shifting Consumer Baskets”; The Food Institute, “Amazon Prime Day’s Quiet Shift: Consumers Are Restocking, Not Splurging”; Nova Analytics, Prime Day 2026 margin and AOV analysis (June 2026); Retail Dive, “Amer Sports CFO: No visibility on tariff refunds” (May 21, 2026)

Thanks for reading. Subscribe for free to receive new posts and support my work.

Subscribe now